# Changelog
Source: https://docs.useraven.ai/changelog
What's new in Raven - the open-source, BYOK meeting copilot for macOS and Windows.
Every release of Raven. Prebuilt installers live on the [download page](https://useraven.ai/download); source and signed assets are on [GitHub Releases](https://github.com/Laxcorp-Research/project-raven/releases). Raven is BYOK - enter your own API keys on first launch.
## macOS in-app self-update fix
* **macOS updates now install.** Earlier builds could download an update, show **Restart & update**, relaunch, and still be on the old version. Bundled libraries shipped read-only, so the macOS updater could not clear the download's quarantine attribute and aborted the install. Libraries are now writable, with a build guard so it cannot regress. If you were stuck on 2.4.1 or 2.4.2, updating to 2.4.3 applies cleanly. See [in-app updates](/features/updates).
* **Updater errors are now logged** instead of silenced, so any future update problem is diagnosable.
macOS only - Windows self-update was unaffected.
## Stealth dock fix and key validation
* **Undetectable no longer shows a Dock icon.** In stealth mode the macOS Dock icon could intermittently reappear when the dashboard was shown; it now stays hidden.
* **API key validation no longer hangs.** Entering an OpenAI, Deepgram, AssemblyAI, or Anthropic key and clicking **Save & Validate** or **Test Connection** could spin forever if the network stalled. Validation now times out with a clear "Could not reach ..." message, so the key can be saved once reachable.
## Streaming Ask, saved chats, and macOS self-update
* **Ask streams now.** Answers type out token-by-token in both per-meeting [Ask](/features/ask) and **Ask across all meetings** - no more waiting for the whole reply. The input sits at the bottom like a normal chat.
* **Ask chats are saved.** Each meeting keeps its own Ask conversation, restored when you reopen the session. **Ask across all meetings** now supports multiple chats - start a **new** one, switch, **rename**, or **delete** - each auto-titled from your first question. A turn is never lost if you navigate away mid-answer; generation finishes in the background and the full answer is saved.
* **Microsoft Teams 1:1 calls** are now detected for [auto-start](/features/meeting-detection), not just scheduled meetings.
* **macOS updates in-app, like Windows.** Now that builds are notarized, click **Update now** to download and **Restart & update** to install - no more grabbing a disk image by hand. (Existing 2.4.0 Macs get 2.4.1 through the installer one last time; in-app [self-update](/features/updates) takes over from 2.4.1 onward.)
## Meeting intelligence, after the call
The biggest post-call release yet. Everything here runs locally - retrieval and embeddings are on-device; only your own model calls leave your machine.
* **Ask your meetings.** A new **Ask** tab on every session answers questions from that call's transcript, and a global **Ask across all meetings** searches your whole history with cited sources. It's conversational with memory - follow-ups build on earlier answers, and older turns are summarized (never dropped) so long chats keep their context.
* **Action items.** Every session now surfaces a structured action-items checklist - task, owner, and deadline when they were spoken.
* **Follow-up email draft.** One click turns a session into a ready-to-send recap email. Drafts are saved with the session.
* **Talk ratio.** See your You-vs-Them speaking share live in the overlay and on each session.
* **Export.** Save any session recap as Markdown or PDF, with optional transcript.
* **Auto-start on meeting detection.** Raven can notice a Zoom, Google Meet, Teams, or Webex meeting and offer to start - set it to prompt, start automatically, or off in Settings. No bot ever joins your call.
* **Smarter notes.** Your mode's note-template instructions now guide the summary, so notes follow the structure you set.
* Onboarding now asks for **permissions before API keys**, and saves your wizard step and keys - Quit & Reopen (e.g. after the macOS Screen Recording restart) resumes where you left off instead of wiping keys.
* The Mac update modal no longer offers a version you already have (same-version GitHub feed no longer shows "Raven X is ready").
* **Send Feedback** (user menu and About) opens a form - no GitHub account required. **Docs** point to [docs.useraven.ai](https://docs.useraven.ai).
* Fixed sessions that stayed **Untitled with a blank summary** after a call. Notes now stream at the model's max output so reasoning can't eat the whole reply, and missing notes are retried on boot.
* Notes generation stays on the fast, cheap models (Haiku / Luna); live assist is unchanged with the full model catalog and effort settings.
* Packaged the tray-icon assets into the Windows and macOS installers so the system-tray icon shows up reliably.
* First open-source, BYOK desktop release for Windows and macOS. Publishing it created the GitHub update feed (`latest.yml` / `latest-mac.yml`) so installed copies update cleanly.
* Reliability: graceful handling of client disconnects during AI streaming, plus internal telemetry and stability fixes.
* Hotfix for a real-time sync endpoint that could return errors on certain requests.
* Refreshed the Anthropic (Claude) and OpenAI model lineups to the current generation.
* Made the interview coaching mode more general-purpose, and hardened the app under load testing.
* Windows installers are now **Authenticode-signed** via the release pipeline, reducing SmartScreen friction.
* Dependency and security cleanup.
* Added **anonymous usage telemetry** with clear in-app links to the Privacy Policy and Terms. First release with real users.
* Observability and reliability improvements.
* Restored automatic GitHub Release creation in the publish pipeline.
* Fixed a fatal crash when the app booted twice on macOS.
* Launched the documentation site at [docs.useraven.ai](https://docs.useraven.ai).
* Enabled **macOS notarization** for smoother Gatekeeper installs, with update-UX polish.
* macOS updates now quit and install correctly, with immediate download feedback.
* Per-account database isolation for safer multi-user and multi-environment use.
* Production-grade update experience: user-initiated downloads, an update button that reflects the real update state, and a macOS auto-update (zip) target.
* Reproducible native echo-cancellation (AEC) build pipeline with versioned artifacts.
* Cache-first profile loading, consolidated IPC error handling, and wired-up session summaries.
* Packaging and version bump.
* First build: a **live transcript tab** in the overlay with auto-scroll, a six-step onboarding with an overlay tour, a redesigned Settings screen, and broad security hardening plus a large automated test suite.
# Raven vs Avoma
Source: https://docs.useraven.ai/compare/avoma
A free, open-source Avoma alternative - a live copilot with local notes, no per-seat platform, and no paywall from day one.
Avoma is a **sales conversation-intelligence** platform: recording, coaching analytics, deal insights, and scheduling for RevOps teams. It's capable, but there's **no free tier** (14-day trial), it's **per recorder seat**, the coaching layer is a **paid add-on**, and it's cloud + bot. Raven is **free**, **open source**, **local**, and helps you **live during** the call.
## At a glance
| | Raven | Avoma |
| ------------------------ | --------------------------------- | ---------------------------------------------------------------------------------------- |
| Live in-call assist | **Yes** | No - analysis after the call |
| Free tier | **Yes** - full app | **No** - 14-day trial |
| Where your meetings live | Local SQLite | Avoma's cloud |
| Price | Free, open source (your own keys) | Startup \$19/seat/mo; Organization \$29; + Conversation Intelligence add-on \$29/seat/mo |
| Open source | Yes | No |
## Price
Avoma has **no permanent free plan**. Base plans run **\$19/seat/mo** (Startup, annual; \$29 monthly) to **\$39/seat/mo** (Enterprise), and the actual coaching/CI features are a **\$29/seat/mo add-on** on top. A small sales team quickly lands north of \$50/seat/mo all-in.
Raven is free; you pay your own AI/transcription usage (a few dollars a month for typical use), with no seat fee and no add-ons.
## What Avoma does better (honestly)
* **Sales CI + RevOps**: deal scorecards, coaching analytics, revenue intelligence, lead routing.
* Built-in **scheduling** and deep CRM sync; free viewer seats for the wider team.
If you run a sales org and want coaching analytics wired into your pipeline, Avoma is built for that.
## When Raven wins
* You don't want to **pay from day one** with no free tier, plus per-seat add-ons.
* You want **live** help in the call and meetings kept **local**.
* You want a **free**, **open-source**, auditable tool.
Free, open source, Mac + Windows. Bring your own keys.
Raven vs the whole category, honestly ranked.
Avoma pricing verified August 2026 (avoma.com/pricing). Vendors change pricing often - confirm before relying on a figure.
# Best AI meeting assistants in 2026
Source: https://docs.useraven.ai/compare/best-ai-meeting-assistants
An honest ranking of the top AI meeting assistants - Raven, Cluely, Fathom, Granola, Otter, Fireflies, tl;dv, Krisp, and Circleback - by price, privacy, and whether they help during the call.
There are a lot of "AI notetakers" and they mostly blur together. The useful way to tell them apart is three questions: **Does it help you *during* the call, or only write it up after? Does a bot join the meeting? And where does your data live?**
Raven is the odd one out because it **combines all four**: it's **open source**, runs **local-first** (recordings stay in a database on your machine), does **live in-call assist**, and is **free** (you bring your own API keys). Live assist itself is no longer unique - Fireflies and Otter added it in 2026 - but no other tool here is free, local, bot-free, and open source at the same time. Below is the honest field, including where other tools genuinely win.
We make one of these tools, so treat this as a competitor's ranking and verify anything that matters. Prices verified August 2026; every vendor changes pricing - check their page before you buy.
## The field at a glance
| Tool | Best for | Starting price (per seat) | Free tier | Live in-call help | Bot-free | Local / open source |
| ---------- | ---------------------------------------------- | --------------------------------------------- | ----------------------- | ----------------- | ------------------- | ------------------- |
| **Raven** | Live help + private notes, one tool everywhere | **Free** (your own API keys) | Yes - full app | **Yes** | Yes | **Yes / Yes** |
| Cluely | Live answers if you'll pay for stealth | \$19.99/mo (Pro); \$149.99/mo w/ invisibility | Yes (limited) | Yes | Yes | No / No |
| Fathom | Individual reps who want a free recap | \$16/user/mo | Yes (5 AI summaries/mo) | No | Bot / bot-free beta | No / No |
| Granola | Calm, beautiful bot-free notes | \$14/user/mo (monthly only) | Yes (last 30 days) | No | Yes | No / No |
| Otter | Budget transcription on one platform | \$8.33/user/mo | Yes (300 min/mo) | Enterprise only | Bot | No / No |
| Fireflies | Enabling other tools / CRM breadth | \$10/user/mo | Yes (limited) | Yes (metered) | Bot | No / No |
| tl;dv | Sales managers coaching reps | \$18/user/mo (Pro); \$59 Business | Yes (10 AI notes) | No | Bot | No / No |
| Krisp | Noise cancellation + light notes | \$8/user/mo | Yes (60 min/day) | No | Yes | No / No |
| Circleback | Post-call automations | \$20.83/user/mo (no free tier) | No (7-day trial) | No | Bot | No / No |
| Jamie | Contractual EU data residency, bot-free | €25/mo (Plus) | Yes (10 mtgs/mo) | No | Yes | No / No |
## How to choose
* **You want help *during* the call** ("what should I say?"): **Raven** or **Cluely** are built around it; **Fireflies** (metered by AI credits) and **Otter** (Enterprise) added live assist in 2026 too. Raven includes screen-share stealth free and stays local; Cluely gates invisibility behind its \$149.99/mo tier, and Fireflies/Otter run in their cloud. See [Raven vs Cluely](/compare/cluely).
* **You just want a clean recap and love bot-free**: **Granola** or **Raven**. Granola's notes UX is excellent, but the free plan only shows your last 30 days and there's no live help. See [Raven vs Granola](/compare/granola).
* **You want the most generous free recap**: **Fathom** - but the free plan caps advanced AI summaries at 5 calls/month. See [Raven vs Fathom](/compare/fathom).
* **You mostly need cheap transcription**: **Otter** or **Fireflies** - watch the 300-min/mo (Otter) and storage/AI-credit caps. See [Raven vs Otter](/compare/otter) and [Raven vs Fireflies](/compare/fireflies).
* **You're a sales manager coaching reps**: **tl;dv** or **Circleback** shine at scorecards and automations (and cost the most). See [Raven vs tl;dv](/compare/tldv) and [Raven vs Circleback](/compare/circleback).
* **Your #1 problem is background noise**: **Krisp** leads on noise cancellation, with notes as a bonus. See [Raven vs Krisp](/compare/krisp).
* **You need contractual EU data residency / GDPR paperwork**: **Jamie** is built for that - or Raven's fully-local model if you'd rather the data never leave your machine. See [Raven vs Jamie](/compare/jamie).
## What the others do better (honestly)
* **Cluely**: a polished mobile app and one bundled subscription.
* **Fathom**: calendar auto-join, CRM sync, clips/playlists, mature team features.
* **Granola**: arguably the nicest notes UX in the category, plus deep integrations/MCP.
* **Otter / Fireflies**: big brands, mobile apps, broad integrations, large cloud archives - and both now offer real-time live assist (Otter on Enterprise; Fireflies metered by credits).
* **tl;dv / Circleback**: real sales coaching (MEDDIC/BANT/SPIN) and post-call automations.
* **Krisp**: best-in-class noise cancellation.
If your job is a RevOps workflow or a coached sales floor, those paid tools earn their price.
## Where Raven wins
* **Live + after-call in one tool** - most of this list only does the after-call half.
* **Free and open source** - no per-seat fee; audit exactly what leaves your machine.
* **Local-first** - transcripts, notes, and the Ask index live in SQLite on your device, not a vendor cloud.
* **BYOK** - you pay Deepgram/AssemblyAI and Anthropic/OpenAI directly for usage (typically a few dollars a month, scaling with how much you record and which model you pick), instead of \$8-\$150/seat.
* **No bot** - nothing joins the call; capture is local on macOS and Windows.
## More head-to-heads
Want the specifics? See the full [feature comparison matrix](/compare/feature-matrix) - capture, privacy, transcription/AI, integrations, and price across all ten tools.
Full side-by-side pages: [Cluely](/compare/cluely) · [Fathom](/compare/fathom) · [Granola](/compare/granola) · [Otter](/compare/otter) · [Fireflies](/compare/fireflies) · [tl;dv](/compare/tldv) · [Krisp](/compare/krisp) · [Jamie](/compare/jamie) · [Circleback](/compare/circleback) · [Read.ai](/compare/read-ai) · [Fellow](/compare/fellow) · [Bluedot](/compare/bluedot) · [Avoma](/compare/avoma) · [Sembly](/compare/sembly) · [Tactiq](/compare/tactiq) · [SalesEQ](/compare/saleseq) · [Gong](/compare/gong) · [Platform built-ins](/compare/platform-built-ins)
## Try Raven
Free, open source, Mac + Windows. Bring your own keys.
Side-by-side pages vs Cluely, Fathom, Granola, and Otter.
Pricing verified August 2026 against each vendor's official pricing page. Most figures are the per-seat rate when billed annually (cheaper than month-to-month); Cluely and Jamie are shown at their standard monthly rate. Vendors change pricing often - confirm before relying on a number.
# Raven vs Bluedot
Source: https://docs.useraven.ai/compare/bluedot
A free, open-source Bluedot alternative - bot-free like Bluedot, but local-first, open source, with live in-call help and no per-seat fee.
Bluedot is a well-liked **bot-free** notetaker that captures through a Chrome extension (plus desktop/mobile). Raven is bot-free too, so this comes down to the rest: Bluedot is **cloud** and **per seat** with a **5-meeting lifetime** free plan; Raven is **local**, **open source**, **free**, and adds **live help during** the call.
## At a glance
| | Raven | Bluedot |
| ------------------------ | --------------------------------- | -------------------------------------------------------------- |
| Bot-free capture | Yes (native Mac + Windows) | Yes (Chrome extension + desktop/mobile) |
| Live in-call assist | **Yes** | No - notes after |
| Free tier | Full app, unlimited local history | **5 meetings lifetime** (1-hour cap) |
| Where your meetings live | Local SQLite | Bluedot's cloud |
| Price | Free, open source (your own keys) | Free (5 lifetime); Basic \$14/user/mo; Pro \$20; Business \$32 |
| Open source | Yes | No |
## Price
Bluedot's free plan is a **5-meeting lifetime** trial. Audio-only unlimited starts at **Basic \$14/user/mo** (annual; \$18 monthly), video at **Pro \$20/user/mo**, and CRM at **Business \$32/user/mo**.
Raven is free with unlimited local history; you pay your own AI/transcription usage (a few dollars a month for typical use), no seat fee.
## What Bluedot does better (honestly)
* **Chrome-based** capture runs anywhere Chrome does (incl. ChromeOS/Linux), plus mobile and Apple Watch.
* 60+ language transcription and clean CRM/email automations on higher tiers.
If you want a browser-native bot-free recorder across many devices, Bluedot is a solid pick.
## When Raven wins
* You want **live help during** the call, not just notes after.
* You want meetings **local** and a tool you can **audit** (open source).
* You want **unlimited** history free, not a 5-meeting lifetime cap, and no per-seat fee.
Free, open source, Mac + Windows. Bring your own keys.
Raven vs the whole category, honestly ranked.
Bluedot pricing verified August 2026 (bluedothq.com/pricing). Vendors change pricing often - confirm before relying on a figure.
# Raven vs Circleback
Source: https://docs.useraven.ai/compare/circleback
A free, open-source Circleback alternative with live in-call help and local storage - no paywall from day one, no per-seat cloud subscription.
Circleback is a polished, automation-heavy notetaker - it turns meetings into CRM updates, tasks, and Slack posts automatically. The catch is there's **no free tier at all** (a 7-day trial, then you pay from day one), it's **per-seat cloud**, and it's after-the-call. Raven is **free** and **open source**, keeps meetings **local**, and helps you **live during** the call.
## At a glance
| | Raven | Circleback |
| ------------------------ | --------------------------------- | ----------------------------------------------------------------------------- |
| Free tier | **Yes** - full app | **No** - 7-day trial, then paid |
| Live in-call assist | **Yes** | No - after the call |
| Where your meetings live | Local SQLite | Circleback's cloud |
| Price | Free, open source (your own keys) | Individual \$20.83/user/mo (annual), \$25 monthly; Team \$25/user/mo (annual) |
| Capture | Local, bot-free (Mac + Windows) | Visible bot (plus botless mode) |
| Open source | Yes | No |
## Price
Circleback has **no permanent free plan** - after a 7-day trial you're a paid seat: **Individual \$20.83/user/mo** (annual; \$25 monthly), **Team \$25/user/mo** (annual; \$30 monthly). So every active user costs from the first week.
Raven is free and open source; you pay your own AI/transcription usage (a few dollars a month for typical use), with no trial clock and no seat fee.
## Where your meetings live
Circleback runs in its cloud (and joins calls with a visible bot by default) to power its automations. Raven keeps transcripts and notes in a **local SQLite database** on your machine, with no bot in the meeting; audio only goes to the providers you chose with your own keys. Open source makes the data path auditable.
## What Circleback does better (honestly)
* **Post-call automations** - auto-update CRM, create tasks, post to Slack; plus email context and MCP.
* Very accurate transcription with speaker names, and a clean product.
If your job is "meetings should automatically update my other tools," Circleback's automation engine is genuinely good - that's what you're paying for.
## When Raven wins
* You don't want to **pay from day one** with no free tier.
* You want meetings **local** and **no bot** in the call.
* You want **live help during** the call, not just automations after.
* You want a **free**, **open-source**, auditable tool with no per-seat fee.
## Switching from Circleback
1. Export anything you want to keep from Circleback.
2. [Download Raven](https://useraven.ai/download), add your keys, enable meeting auto-start.
3. Run Raven through your trial window; keep Raven instead of converting to a paid seat. (CRM push is on Raven's roadmap; today, copy the follow-up email/action items out.)
Free, open source, Mac + Windows. Bring your own keys.
Action items, follow-up email, export.
Circleback pricing verified August 2026 (circleback.ai); no free tier, 7-day trial only. Vendors change pricing often - confirm before relying on a figure.
# Raven vs Cluely
Source: https://docs.useraven.ai/compare/cluely
A free, open-source Cluely alternative. Screen-share invisibility is included, not a \$149.99/mo tier - and your meetings stay on your machine.
Cluely and Raven are the two tools that actually help you **live, during a call** - real-time answers in an overlay the people you're meeting with can't see. The difference is what that costs and where your data goes: Cluely puts its signature invisibility behind a **\$149.99/mo** tier and stores your meetings in its cloud; Raven ships stealth **free**, keeps everything in a **local database**, and is **open source**.
## At a glance
| | Raven | Cluely |
| ------------------------- | ------------------------------------- | ------------------------------------------------------------------------------------- |
| Live in-call assist | Yes | Yes |
| Invisible to screen share | **Included, free** | Only on **Pro + Undetectability, \$149.99/mo** |
| Price | Free, open source (your own API keys) | Free (limited); Pro \$19.99/mo; Pro + Undetectability \$149.99/mo (\$44.99/mo annual) |
| Where your meetings live | Local SQLite on your machine | Cluely's cloud account (past meetings stored server-side) |
| Bring your own AI keys | Yes - you pay providers directly | No - bundled into the subscription |
| Open source | Yes - auditable | No - closed |
| Platforms | macOS + Windows (native capture) | macOS + Windows; mobile app |
## Price
Cluely's headline feature - the overlay staying hidden during screen share - is only on the **\$149.99/mo** Pro + Undetectability plan (\$44.99/mo if you pay a year up front, i.e. **\$539.88/year**). Pro without invisibility is \$19.99/mo.
Raven includes stealth for **\$0**. You bring your own API keys, so you pay Deepgram/AssemblyAI and Anthropic/OpenAI directly for usage - typically a few dollars a month, and nothing when you're idle. No per-seat fee, no tier gating the core feature.
## Where your meetings live
Cluely is a cloud product: you create an account, and your notes and past-meeting history live on its servers (its free tier even advertises "Ask AI about all your past meetings"). Raven has **no account and no server** - transcripts, notes, and the Ask index sit in a local SQLite database on your machine. Audio goes only to the transcription/AI providers **you** chose with your **own** keys. Because Raven is open source, "it's private" is something you can verify in the code, not just a promise.
## What Cluely does better (honestly)
* A polished **mobile app** and a very refined onboarding.
* One bundled subscription - no thinking about API keys or provider billing.
If you want a turnkey hosted product and don't mind the cloud or the price, Cluely is well-built.
## When Raven wins
* You don't want to pay **\$149.99/mo** for invisibility that's free here.
* You want meetings on **your machine**, not a vendor's cloud.
* You want to **audit** the tool that sits in your calls (open source).
* You'd rather pay your AI provider **cents** than a flat subscription.
## Switching from Cluely
1. [Download Raven](https://useraven.ai/download) (Mac or Windows) and add your API keys on first launch.
2. Run both during your next few calls to compare the live assist and the notes.
3. Cancel Cluely once Raven covers your daily habit. (History importers are on the roadmap; today, export anything you need from Cluely first.)
Free, open source, Mac + Windows. Bring your own keys.
Audit exactly what leaves your machine.
Cluely pricing verified August 2026 (cluely.com/pricing). Vendors change pricing often - confirm before relying on a figure.
# Raven vs Fathom
Source: https://docs.useraven.ai/compare/fathom
A free, open-source Fathom alternative with unlimited AI notes on your own key, live in-call help, and no bot - recordings stay on your machine.
Fathom is a great free notetaker - until you hit the part most people miss: on the free plan, **advanced AI summaries and action items are capped at 5 calls a month**. After that you get a basic transcript-style summary until the cycle resets. Raven gives you **unlimited** AI notes because they run on **your own API key**, adds **live help during the call** (Fathom is after-the-call only), and keeps everything **local** with **no bot** in the meeting.
## At a glance
| | Raven | Fathom |
| --------------------------- | -------------------------------------- | ------------------------------------------------------------------------------------ |
| Live in-call assist | **Yes** - overlay suggests what to say | No - recap is after the call |
| AI summaries / action items | **Unlimited** (your own key) | Free: **5 calls/month**, then basic only; unlimited on paid |
| Follow-up email draft | Yes | Yes (paid) |
| Bot in the meeting | **Never** - local capture | Bot, or bot-free capture in beta (Mac) |
| Where your meetings live | Local SQLite | Fathom's cloud |
| Price | Free, open source (your own keys) | Free (capped AI); Premium \$16/user/mo (annual), \$20 monthly; Business \$25/user/mo |
| Open source | Yes | No |
| Platforms | macOS + Windows (native capture) | Zoom-native; Meet/Teams; Mac bot-free beta |
## Price
Fathom's free tier is genuinely generous on **recording, transcription, and storage** (no minute cap) - the limit is the AI: **5 advanced summaries/action-item runs per month**. To remove that you're on **Premium at \$16/user/mo** (billed annually; \$20 monthly), and CRM sync lives in **Business at \$25/user/mo**.
Raven's AI is **unlimited** because you're billing your own provider for usage - typically a few dollars a month - with **no per-seat fee**.
## Where your meetings live
Fathom stores your recordings, transcripts, and summaries in its cloud. Raven keeps them in a **local SQLite database** on your machine - no account, no server, and (because it's open source) a data path you can audit. Audio only ever goes to the transcription/AI providers you configured with your own keys.
## What Fathom does better (honestly)
* **Calendar auto-join**: Fathom's bot joins scheduled meetings automatically. Raven detects a meeting and offers to start (or auto-starts) locally, but it doesn't join from your calendar.
* **CRM sync** (HubSpot/Salesforce), **clips/playlists**, **team collaboration**, and **15+ summary templates** are mature.
* A very polished, zero-setup hosted experience.
If you're a team that lives in a CRM and wants shared clips, Fathom's paid tiers are strong.
## When Raven wins
* You want **unlimited** AI notes without a 5-a-month cap or a subscription.
* You want **help during** the call, not just a recap after.
* You don't want a **bot** joining, and you want recordings **on your machine**.
* You're on **Windows** and want native capture (not just Mac bot-free beta).
## Switching from Fathom
1. [Download Raven](https://useraven.ai/download) and add your keys.
2. Set meeting auto-start (Settings → prompt or auto) so you never forget to record.
3. Run both for a week; compare notes + action items. Cancel Fathom once Raven owns the habit. (Fathom importers are on the roadmap; export what you need from Fathom in the meantime.)
Free, open source, Mac + Windows. Bring your own keys.
Structured action items, follow-up email, export.
Fathom pricing and the 5-call free AI cap verified August 2026 (fathom.ai/pricing). Vendors change pricing often - confirm before relying on a figure.
# Feature comparison matrix
Source: https://docs.useraven.ai/compare/feature-matrix
A deep, honest feature-by-feature comparison of Raven vs Cluely, Fathom, Granola, Otter, Fireflies, tl;dv, Krisp, Circleback, and Jamie - capture, privacy, AI, integrations, and price.
This is the long-form matrix for people who want to compare on specifics, not vibes. For a shorter, opinionated ranking see [Best AI meeting assistants](/compare/best-ai-meeting-assistants); for one-on-one pages see the [comparison overview](/compare/overview).
We make Raven, so treat this as a competitor's matrix and verify anything that matters before you buy. Facts checked **August 2026** against vendors' own pages and app listings. A dash (-) means "not a focus, or not publicly documented" - not necessarily "impossible." Language counts are vendor-claimed. Prices are per seat; most vendors are cheaper billed annually than monthly.
## What "live in-call assist" means here
The category shifted in 2026, so this row needs a precise definition. **Live in-call assist** means real-time suggestions or answers that help you *respond during the call* (a private panel with talking points, objection help, or mid-call Q\&A) - not just a live transcript or live notes.
* **Raven** and **Cluely** were built around it. Raven's is **free, local, bot-free, and hidden from screen share**; Cluely gates screen-share invisibility behind its \$149.99/mo tier.
* **Fireflies** added **Live Assist** (real-time suggestions + mid-call AskFred answers + Sales Assist), but advanced use is **metered by AI credits** and runs in its cloud.
* **Otter** launched **Live Assist** (a real-time coaching agent) in July 2026, but it is **Enterprise-only**; its live AI Chat is available more broadly on paid tiers.
* **Fathom** shows **live notes/summaries** during the call but does not suggest what to say. Granola, tl;dv, Krisp, Circleback, and Jamie are after-the-call.
So Raven is no longer the only tool with live help - its distinction is that live help is **free, on your machine, without a bot, and open source**.
## Help and capture
| Tool | Live in-call assist | Bot in the call? | Runs on |
| ---------- | ---------------------------------------- | ------------------------ | ---------------------------------- |
| **Raven** | **Yes** - free, hidden from screen share | **No** - local capture | Mac, Windows |
| Cluely | Yes - invisibility on \$149.99/mo | No - local overlay | Mac, Windows, mobile |
| Fathom | Live notes only | Bot; bot-free beta (Mac) | Mac, Windows (no mobile yet) |
| Granola | No - notes only | No - local capture | Mac, Windows, iOS, Android |
| Otter | Enterprise only (+ live AI chat) | Bot | Web, desktop, iOS, Android |
| Fireflies | Yes - metered by AI credits | Bot (+ no-bot mode) | Web, desktop, Chrome, iOS, Android |
| tl;dv | No - coaching is after the call | Bot or bot-free | Mac, Windows, iOS |
| Krisp | No | No - local capture | Mac, Windows, mobile |
| Circleback | No - automations after the call | Bot (+ botless) | Web, Mac, iOS, Android |
| Jamie | No - post-meeting | No - local capture | Mac, Windows, iOS |
## Privacy and data
| Tool | Where your data lives | Open source | History / retention control |
| ---------- | -------------------------------------------------- | ------------- | ------------------------------------------------------ |
| **Raven** | **Local SQLite on your device** | **Yes (MIT)** | Unlimited local history; you delete; incognito + reset |
| Cluely | Cluely cloud | No | Cloud account |
| Fathom | Fathom cloud | No | Cloud; free storage is unlimited |
| Granola | Granola cloud | No | Cloud; free plan shows only last 30 days |
| Otter | Otter cloud | No | Cloud; free 300 min/mo |
| Fireflies | Fireflies cloud | No | Cloud; free storage is limited |
| tl;dv | tl;dv cloud (EU-hosted) | No | Cloud; free auto-deletes after 3 months |
| Krisp | Krisp cloud | No | Cloud; free keeps 7-day history |
| Circleback | Circleback cloud | No | Cloud |
| Jamie | Jamie EU cloud (audio deleted after transcription) | No | Cloud |
## Transcription and AI
| Tool | Speaker attribution | Languages | AI notes / action items | Ask across meetings |
| ---------- | ------------------------------ | ---------------------------- | ----------------------------------------- | ------------------------------ |
| **Raven** | You / Them, on device | Many (via your STT provider) | **Unlimited** (your own key) | **Yes** - local, cited sources |
| Cluely | Basic | Multiple | Yes (bundled) | Yes (cloud) |
| Fathom | Named speakers | 20+ (claimed) | Free: 5/mo, then basic; unlimited on paid | Ask Fathom (cloud) |
| Granola | Named speakers | Multiple | Yes | Granola chat (cloud) |
| Otter | Named speakers (voiceprints) | Multiple (English-first) | Yes | Otter AI Chat (cloud) |
| Fireflies | Named speakers | 100+ | Metered by AI credits | AskFred (uses credits) |
| tl;dv | Named speakers | 30+ (claimed) | Free: 10 notes | Ask tl;dv (cloud) |
| Krisp | Basic | Multiple | 2 notes/day free | Limited |
| Circleback | Named speakers (auto voice ID) | 100+ | Yes | Yes (cloud, cited) |
| Jamie | Named speakers | 100+ | Yes | Ask Jamie (cloud) |
## Team and extensibility
| Tool | Integrations / CRM | API / MCP | Admin / SSO | Export |
| ---------- | ----------------------------------------------------------------------- | ------------------------------------- | --------------------------- | ------------------------- |
| **Raven** | None built-in (export + copy); on roadmap | Open source - self-host and script it | No - single-user local app | Markdown, PDF, copy |
| Cluely | Minimal | - | - | Copy |
| Fathom | Slack, Salesforce, HubSpot, Notion, Asana | MCP (ChatGPT/Claude) | Business / Enterprise | PDF, email, CRM |
| Granola | HubSpot, Notion, Slack, Attio, Zapier | MCP + API (Business) | Enterprise | Copy, integrations |
| Otter | Slack, Salesforce, HubSpot, and more | API | Business / Enterprise (SSO) | PDF, TXT, integrations |
| Fireflies | 100+ apps, deep CRM | API | Business / Enterprise (SSO) | Integrations, API |
| tl;dv | HubSpot, Salesforce, Slack | API | Business (SSO) | CRM, integrations |
| Krisp | Limited | - | Enterprise | Copy |
| Circleback | Salesforce, HubSpot, Attio, Zoho, Slack, Notion, Linear, monday, Zapier | MCP, Zapier | Team / Enterprise | Integrations, automations |
| Jamie | Integrations (Pro) | API | SSO (Team) | PDF, integrations |
## Pricing and free tier
| Tool | Free tier | Free-tier limit | Starting paid (per seat) | Notes |
| ---------- | ------------------ | -------------------------------------- | ---------------------------------------- | ------------------------------------------------------------------- |
| **Raven** | **Yes - full app** | None (you pay your own API usage) | **Free** (bring your own keys) | Open source; API usage is typically a few \$/mo and scales with use |
| Cluely | Yes (limited) | Limited | Pro \$19.99/mo; invisibility \$149.99/mo | Screen-share invisibility gated to the top tier |
| Fathom | Yes | 5 AI summaries/mo | Premium \$16/user/mo (annual) | Business \$25 adds CRM sync |
| Granola | Yes | Last 30 days of history | Business \$14/user/mo (monthly only) | No annual discount |
| Otter | Yes | 300 min/mo, 30-min/conversation | Pro \$8.33/user/mo (annual) | Business \$19.99 unlimited; Live Assist is Enterprise |
| Fireflies | Yes | Limited storage + AI credits | Pro \$10/user/mo (annual) | Business \$19 |
| tl;dv | Yes | 10 AI notes; auto-delete after 3 mo | Pro \$18/user/mo (annual) | Business \$59 adds sales coaching |
| Krisp | Yes | 60 min/day, 2 notes/day, 7-day history | Core \$8/user/mo (annual) | Noise cancellation is the core product |
| Circleback | No - 7-day trial | - | Individual \$20.83/user/mo (annual) | Team \$25/user/mo |
| Jamie | Yes | 10 meetings/mo, 30-min cap | Plus €25/mo | Pro €47; Team €39/seat; billed in EUR, annual lower |
## How to read this
* **If you want live help during the call**: Raven, Cluely, Fireflies, or Otter (Enterprise). Only Raven is free, local, bot-free, and open source.
* **If privacy or data ownership is the point**: Raven (local + open source) or Jamie (managed EU cloud). Everyone else is closed cloud.
* **If you need breadth** - 100+ integrations, mobile apps, SSO, big archives: Fireflies, Otter, or Circleback earn their price. Raven is deliberately a single-user, local desktop tool.
* **If cost is the driver**: Raven is free software (you pay only your own API usage); the rest are \$8-\$150 per seat per month.
Free, open source, Mac + Windows. Bring your own keys.
The field, ranked - including where others win.
Every figure here was verified August 2026 against each vendor's official pricing page, docs, or app-store listing. Vendors change features and pricing often - confirm before relying on a specific number.
# Raven vs Fellow
Source: https://docs.useraven.ai/compare/fellow
A free, open-source alternative to Fellow - a live copilot + private notes for the individual, not a team meeting-governance suite.
Fellow is really a **team meeting-management** platform - shared agendas, recurring 1-on-1s, action-item tracking, admin controls and governance. It's good at that. But it's built for the org, it's **per seat**, its free plan caps AI at **5 notes lifetime**, and it doesn't help you **during** the call. Raven is for the **individual in the meeting**: live help, private notes you own, free and open source.
## At a glance
| | Raven | Fellow |
| ------------------------ | --------------------------------- | ----------------------------------------------------------------------- |
| Built for | The individual in the call | Teams / meeting governance |
| Live in-call assist | **Yes** | No |
| Free-tier AI | Unlimited (your own key) | **5 AI notes + 5 recordings, lifetime** |
| Where your meetings live | Local SQLite | Fellow's cloud |
| Price | Free, open source (your own keys) | Free (capped); Team \$7/user/mo; Business \$15/user/mo; Enterprise \$25 |
| Open source | Yes | No |
## Price
Fellow's free tier is a trial - **5 AI notes and 5 recordings per user, lifetime**. Real use is **Team \$7/user/mo** (annual; \$11 monthly, 10 AI notes/mo) or **Business \$15/user/mo** for unlimited AI. Every tier is per seat.
Raven is free with unlimited AI on your own key - no seat fee, no lifetime cap.
## What Fellow does better (honestly)
* **Meeting governance**: collaborative agendas, structured 1-on-1s, feedback loops, org-wide templates.
* Admin controls, SSO, HRIS sync, analytics - real value for managers and ops teams.
If your goal is fixing a team's meeting culture with agendas and accountability, Fellow is purpose-built and Raven isn't trying to be that.
## When Raven wins
* You're an **individual** who wants live help and a private recap - not a governance committee.
* You want meetings **local** and a **free**, auditable tool with no per-seat fee.
* You want **help during** the call.
Free, open source, Mac + Windows. Bring your own keys.
Raven vs the whole category, honestly ranked.
Fellow pricing verified August 2026 (fellow\.ai/pricing). Vendors change pricing often - confirm before relying on a figure.
# Raven vs Fireflies
Source: https://docs.useraven.ai/compare/fireflies
A free, open-source Fireflies alternative - local-first, no per-seat fee or AI credits, with live in-call help. Your meetings stay on your machine.
Fireflies is a deep, integration-heavy notetaker built around a cloud archive of your calls. If you want 100+ integrations and a big searchable cloud, it's strong. But it's **per-seat**, meters advanced AI behind **AI credits**, and everything lives in **Fireflies' cloud**. Raven keeps meetings in a **local database**, has **no seat fee or credits**, adds **live help during the call** (Fireflies has Live Assist too, but it draws down AI credits), and is **open source**.
## At a glance
| | Raven | Fireflies |
| ------------------------ | --------------------------------- | ------------------------------------------------------------------------------------------------ |
| Live in-call assist | **Yes** - free, unmetered | Yes - Live Assist, but metered by AI credits |
| Where your meetings live | Local SQLite | Fireflies' cloud archive |
| Advanced AI | **Unlimited** (your own key) | Metered by **AI credits** on top of the seat |
| Price | Free, open source (your own keys) | Free (limited storage + credits); Pro \$10/user/mo (annual), \$18 monthly; Business \$19/user/mo |
| Capture | Local, bot-free (Mac + Windows) | Notetaker bot (plus some browser/no-bot modes) |
| Open source | Yes | No |
## Price
Fireflies' free tier records and transcribes, but caps **storage** and hands you a small one-time pool of **AI credits** - the advanced summaries and AskFred draw them down, so heavy use pushes you to **Pro at \$10/user/mo** (annual; \$18 monthly) and integrations/CRM depth to **Business at \$19/user/mo**. Every tier is per seat.
Raven's AI is unlimited because you bill your own provider for usage - typically a few dollars a month - with no seat fee and no credit system.
## Where your meetings live
Fireflies stores your recordings and transcripts in its cloud so it can offer archive-wide search and integrations. Raven keeps everything in a **local SQLite database** on your machine - no account, no server - and its Ask feature searches your history from a **local** index. Audio only goes to the providers you chose with your own keys, and because Raven is open source, that path is auditable.
## What Fireflies does better (honestly)
* **Integration breadth** - 100+ apps, deep CRM (HubSpot/Salesforce/Zoho/Pipedrive), and AskFred.
* **Live Assist** - real-time suggestions and mid-call AskFred answers (metered by AI credits).
* A big **cloud archive** with cross-meeting search and analytics on higher tiers.
* Mature mobile apps and team features.
If your workflow is "route meeting data into a dozen other tools," Fireflies is built for that.
## When Raven wins
* You don't want a **per-seat** subscription or **AI-credit** metering.
* You want meetings **on your machine**, not a cloud archive.
* You want **live help that's free and unmetered** - Fireflies has Live Assist, but it draws down AI credits.
* You want to **audit** the tool (open source) and pay your AI provider **cents**.
## Switching from Fireflies
1. Export any transcripts you want to keep from Fireflies.
2. [Download Raven](https://useraven.ai/download), add your keys, enable meeting auto-start.
3. Run both for a week; cancel Fireflies once Raven covers your notes habit. (Importers are on the roadmap.)
Free, open source, Mac + Windows. Bring your own keys.
Raven vs the whole category, honestly ranked.
Fireflies pricing verified August 2026 (fireflies.ai/pricing). Vendors change pricing often - confirm before relying on a figure.
# Raven vs Gong
Source: https://docs.useraven.ai/compare/gong
Conversation intelligence without the enterprise contract - a free, open-source, local alternative to Gong for the individual seller, not the RevOps org.
Gong is the enterprise **revenue-intelligence** standard: deal boards, forecasting, org-wide coaching. It's also an enterprise **purchase** - no public price, no free tier, a mandatory platform fee, and multi-year contracts. Raven doesn't pretend to be a RevOps platform. It's the answer to a different question: **can I get live help and honest call insight as an individual, without a five-figure contract?** Yes - free, open source, and local.
## At a glance
| | Raven | Gong |
| --------------------- | --------------------------------- | -------------------------------------------------------------------------------------- |
| Built for | The **individual seller** / IC | Enterprise revenue teams |
| Live in-call assist | **Yes** | No - post-call revenue intelligence |
| Free tier | **Yes** | No |
| Commitment | None - download and go | Annual/multi-year + platform fee + onboarding |
| Price | Free, open source (your own keys) | No public price: \~\$1,300-\$1,920/seat/yr **plus** a \$5,000-\$50,000/yr platform fee |
| Where your data lives | Local SQLite | Gong's cloud |
| Open source | Yes | No |
## Price
Gong doesn't publish pricing. Third-party buyer data puts the Foundations tier around **\$1,300-\$1,920 per seat per year**, on top of a **mandatory platform fee of \$5,000-\$50,000/year** and one-time onboarding (often \$7,500+). Effective cost lands near **\$2,100+/seat/year**, on 2-3 year contracts, with no free tier.
Raven is free; you pay your own AI/transcription usage (a few dollars a month for typical use) - no contract, no platform fee.
## What Gong does better (honestly)
* **Enterprise revenue intelligence**: pipeline/deal analytics, forecasting, org-wide coaching and analytics at team scale.
* A mature platform with deep CRM and enterprise controls.
If you're equipping a sales org with revenue intelligence and forecasting, Gong is a category leader - Raven is **not** a replacement for that deployment, and won't pretend to be.
## When Raven wins
* You're an **individual or small team** who wants **live help + private notes + talk ratio** without a five-figure annual contract.
* You want meetings kept **local**, no bot, and a tool you can **audit**.
* You want it **free** - conversation intelligence without the enterprise procurement.
Free, open source, Mac + Windows. Bring your own keys.
Raven vs the whole category, honestly ranked.
Gong does not publish pricing; figures are third-party buyer-data estimates verified August 2026 (Vendr and others). Confirm any number directly.
# Raven vs Granola
Source: https://docs.useraven.ai/compare/granola
A free, open-source Granola alternative with live in-call assist and unlimited local history - no 30-day free-tier lock, no cloud, no per-seat fee.
Granola proved people love **bot-free** notes, and its writing is excellent. Two things it doesn't do: it has **no live in-call assist**, and on the free plan you can only open the **last 30 days** of your notes - older ones are stored but locked until you pay **\$14/user/mo**. Raven is bot-free too, but adds **real-time help during the call**, keeps **unlimited history locally and free**, and is **open source**.
## At a glance
| | Raven | Granola |
| ----------------------------- | --------------------------------- | ------------------------------------------------------------------------------------ |
| Live in-call assist | **Yes** | No - notes only |
| Bot-free capture | Yes (Mac + Windows) | Yes (Mac + Windows) |
| Note history on the free plan | **Unlimited, local** | **Last 30 days** in-app (older stored but locked) |
| Where your meetings live | Local SQLite | Granola's cloud |
| Price | Free, open source (your own keys) | Free (30-day history); Business **\$14/user/mo, monthly only**; Enterprise from \$35 |
| Open source | Yes | No |
| Annual discount | n/a (free) | None - monthly billing only |
## Price
Granola's free **Basic** plan is a real product, but its catch is time: **only the last 30 days** of notes are accessible in the app. To unlock full history (the main reason people upgrade) you're on **Business at \$14/user/mo** - and Granola bills **monthly only**, with **no annual discount**, so that's \~\$168/user/year.
Raven keeps **all** your history, locally, for **free**. Your only cost is your own API usage (a few dollars a month for regular use).
## Where your meetings live
Both are calm and bot-free. The difference is ownership: Granola stores your notes in its cloud (older ones remain there, just gated on the free plan). Raven stores everything in a **local SQLite database** you own - no account, no server - and being open source, that's auditable rather than promised.
## What Granola does better (honestly)
* A beautifully refined **notes UX** and template/recipe system.
* Mature **integrations** (HubSpot, Notion, Slack, Attio, Zapier) and MCP/API access on Business.
* Native **mobile apps** (iOS/Android) and cross-device sync; Raven is desktop-only (Mac + Windows).
* Team workspace features.
If you want a polished hosted notes app with deep integrations and don't need live help, Granola is lovely.
## When Raven wins
* You want **help during** the call, not just a clean write-up after.
* You want **all** your history, **free**, without a 30-day wall.
* You want meetings **on your machine** and a tool you can **audit**.
* You'd rather pay your AI provider **cents of usage** than **\$14/seat/mo** to unlock full history.
## Switching from Granola
1. In Granola, export a CSV of your notes (Settings → Profile → Generate CSV) so you keep your back catalog.
2. [Download Raven](https://useraven.ai/download), add your keys, and turn on meeting auto-start.
3. Run both for a week; cancel Granola once Raven covers your notes habit. (Importers are on the roadmap.)
Free, open source, Mac + Windows. Bring your own keys.
Local Q\&A over your whole history, with cited sources.
Granola pricing and the 30-day free-history limit verified August 2026 (granola.ai/pricing, docs.granola.ai). Vendors change pricing often - confirm before relying on a figure.
# Raven vs Jamie
Source: https://docs.useraven.ai/compare/jamie
A free, open-source alternative to Jamie - auditable, fully-local privacy plus live in-call help, vs Jamie's closed, EU-cloud, post-meeting notes.
Jamie is a serious **privacy-first, bot-free** notetaker built in Germany, with genuine EU data residency and clean compliance credentials. It competes for exactly the buyer Raven does: someone who cares where their meetings go. The difference is the **kind** of privacy. Jamie asks you to **trust** a closed product that uploads audio to its EU cloud; Raven is **open source** and keeps everything **on your machine** - privacy you can **audit**, not just be promised - and it also helps you **live during** the call.
## At a glance
| | Raven | Jamie |
| ------------------------ | ---------------------------------- | ---------------------------------------------------------------------------- |
| Live in-call assist | **Yes** | No - post-meeting only |
| Where your meetings live | **Local SQLite** on your device | Uploaded to Jamie's EU cloud (audio deleted after transcription) |
| Privacy model | **Open source** - auditable, local | Closed source - trusted, EU-hosted, ISO 27001 / GDPR |
| Bot-free capture | Yes (Mac + Windows) | Yes (Mac + Windows + iOS) |
| Price | Free, open source (your own keys) | Free (10 meetings/mo, 30-min cap); Plus €25/mo; Pro €47/mo; Team €39/user/mo |
| Open source | Yes | No |
## Price
Jamie's free tier is a real evaluation tier - **10 meetings/month, 30-minute cap** - with paid plans (in EUR) at **Plus €25/mo** (20 meetings, 2-hr cap), **Pro €47/mo** (unlimited, 3-hr cap, integrations), and **Team €39/user/mo**.
Raven is free and open source; you pay your own AI/transcription usage (a few dollars a month for typical use), with no meeting cap and no seat fee.
## Two kinds of privacy
Both are bot-free, which already beats the cloud-bot notetakers on privacy. The distinction:
* **Jamie**: audio is captured on your device, then **uploaded to Jamie's EU infrastructure** for transcription/summaries (and deleted after). Strong, compliance-grade - but it's a **closed** system you have to trust, and your content does leave your machine.
* **Raven**: transcripts, notes, and the Ask index stay in a **local SQLite database**; audio goes only to the transcription/AI providers **you** chose with **your own** keys. Because Raven is **open source**, "it's private" is verifiable in the code.
## What Jamie does better (honestly)
* **Managed EU data residency**, ISO 27001, GDPR/DORA, DPAs, SSO - turnkey compliance for regulated European teams.
* A polished product with structured notes, "Ask Jamie" across meetings, and 100+ languages, including iOS.
If you need a vendor that contractually guarantees EU residency and compliance paperwork, Jamie is purpose-built - Raven's model is "you hold the data," which is a different (also strong) answer to privacy.
## When Raven wins
* You want privacy you can **audit** (open source) and data that **never leaves your machine** except your own API calls.
* You want **live help during** the call, not just notes after.
* You want it **free**, with no meeting cap or per-seat fee.
## Switching from Jamie
1. Export any notes you want to keep from Jamie.
2. [Download Raven](https://useraven.ai/download), add your keys.
3. Run both for a week; cancel Jamie once Raven covers your habit.
Free, open source, Mac + Windows. Bring your own keys.
Local SQLite, incognito, and reset.
Jamie pricing verified August 2026 (meetjamie.ai); Jamie bills in EUR and annual rates are lower. Vendors change pricing often - confirm before relying on a figure.
# Raven vs Krisp
Source: https://docs.useraven.ai/compare/krisp
A free, open-source alternative to Krisp's meeting notes - with live in-call help, local storage, and no per-seat fee. (Krisp still wins on noise cancellation.)
Krisp is best known for **noise cancellation** - genuinely the best in the category - and bolts on bot-free meeting notes. Raven isn't a noise-cancellation product; it's a **live copilot + local notetaker**. So this comes down to what you actually need: cleaner audio, or help during the call plus private notes you own.
## At a glance
| | Raven | Krisp |
| ------------------------ | --------------------------------- | ------------------------------------------------------------ |
| Core strength | Live in-call assist + local notes | **Noise cancellation** (+ notes) |
| Live in-call assist | **Yes** | No |
| Notes free-tier limit | Full app, unlimited local history | 60 min/day noise cancellation, 2 AI notes/day, 7-day history |
| Where your meetings live | Local SQLite | Krisp's cloud |
| Price | Free, open source (your own keys) | Free (metered); Core \$8/user/mo (annual), \$16 monthly |
| Open source | Yes | No |
## Price
Krisp's free tier gives 60 minutes/day of noise cancellation, unlimited transcription, and **2 AI notes/day** with **7-day history**. Unlimited noise cancellation and notes start at **Core, \$8/user/mo** (annual; \$16 monthly).
Raven is free and open source; you pay your own AI/transcription usage (a few dollars a month for typical use), with no per-seat fee.
## Where your meetings live
Krisp processes noise cancellation locally but stores notes/history in its cloud, and the free plan only keeps **7 days** of history. Raven keeps **all** history in a **local SQLite database** on your machine, indefinitely, and audio only goes to the providers you chose with your own keys.
## What Krisp does better (honestly)
* **Noise cancellation** - category-leading, bidirectional, and it works across every app. This is a real, distinct value Raven does not offer as a product.
* Accent conversion and audio features on higher tiers.
If your #1 problem is background noise on calls, Krisp is the right tool - and it's cheap.
## When Raven wins
* You want **help during** the call, not just clean audio and a summary.
* You want **unlimited local history**, not a 7-day free window in someone's cloud.
* You want a **free**, **open-source**, auditable tool with no seat fee.
* Full post-call intelligence: [action items, follow-up email, export](/features/meeting-intelligence), and [Ask across your meetings](/features/ask).
## Switching from Krisp
1. [Download Raven](https://useraven.ai/download) and add your keys.
2. Keep Krisp for noise cancellation if you love it - Raven doesn't replace that; it replaces the notetaking + adds live help.
3. Move your notes habit to Raven.
Free, open source, Mac + Windows. Bring your own keys.
Raven vs the whole category, honestly ranked.
Krisp pricing verified August 2026 (krisp.ai/pricing). Vendors change pricing often - confirm before relying on a figure.
# Raven vs Otter
Source: https://docs.useraven.ai/compare/otter
A free, open-source Otter.ai alternative - no 300-minute cap, no bot, no cloud, plus live in-call assist. Recordings stay on your machine.
Otter is the legacy brand in meeting transcription, but its free plan is tightly metered - **300 minutes a month, a 30-minute cap per conversation, and 3 lifetime file imports** - it captures with a **bot**, stores everything in its **cloud**, and reserves its real-time **Live Assist** coaching for **Enterprise**. Raven has **no minute cap**, captures **locally with no bot**, keeps recordings **on your machine**, adds **free real-time assist** on every tier, and is **open source**.
## At a glance
| | Raven | Otter |
| ------------------------ | ----------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| Live in-call assist | **Yes** - free | Live Assist (Enterprise) + live AI chat |
| Free-tier limit | **No minute cap** | **300 min/month**, 30-min per conversation, 3 lifetime imports |
| Capture | Local, **bot-free** (Mac + Windows) | Bot joins the meeting |
| Where your meetings live | Local SQLite | Otter's cloud |
| Price | Free, open source (your own keys) | Free (300 min); Pro \$8.33/user/mo annual (\$16.99 monthly); Business \$19.99/user/mo annual (\$30 monthly) |
| Open source | Yes | No |
## Price
Otter's free tier is really a trial: **300 transcription minutes/month**, **30 minutes max per conversation**, and **3 lifetime imports**. Pro (\$8.33/user/mo annual, \$16.99 monthly) raises that to 1,200 minutes; unlimited transcription only arrives on **Business (\$19.99/user/mo annual, \$30 monthly)**.
Raven doesn't meter minutes at all. You pay your own transcription/AI provider for usage - typically a few dollars a month - with no per-seat fee.
## Where your meetings live
Otter records with a **bot** that joins your call, and stores transcripts in its **cloud**. Raven captures **locally** (macOS ScreenCaptureKit + CoreAudio, Windows WASAPI), splits **You** and **Them** on-device, and writes everything to a **local SQLite database**. No bot shows up in the meeting; audio only goes to the providers you chose with your own keys. Open source means the data path is auditable.
## What Otter does better (honestly)
* A big, mature brand with polished **iOS/Android apps**.
* Deep **integrations** and meeting-wide search across a large archive.
* **Live Assist** real-time coaching (Enterprise) and live AI Chat during meetings.
* OtterPilot and team/admin features for larger orgs.
If you need mature mobile apps and a big cloud archive with team admin, Otter is established.
## When Raven wins
* You keep hitting the **300-minute** or **30-minute-per-call** free limits.
* You don't want a **bot** in the meeting or your transcripts in a **cloud**.
* You want **free live help during** the call - Otter's Live Assist is Enterprise-only.
* You'd rather pay your AI provider **cents** than a per-seat subscription - and audit the tool.
## Switching from Otter
1. Export any transcripts you want to keep from Otter.
2. [Download Raven](https://useraven.ai/download), add your keys, enable meeting auto-start.
3. Run both for a week; cancel Otter once Raven covers your transcription + notes habit. (Importers are on the roadmap.)
Free, open source, Mac + Windows. Bring your own keys.
Dual-stream local capture, You/Them, your own STT key.
Otter pricing and free-tier limits verified August 2026 (otter.ai/pricing). Vendors change pricing often - confirm before relying on a figure.
# Raven vs the alternatives
Source: https://docs.useraven.ai/compare/overview
How Raven compares to Cluely, Fathom, Granola, and Otter - open-source, local-first, with live in-call assist and no bot.
Most meeting tools are built to **write up the call afterward** (Fathom, Granola, Otter). A few now add **live in-call help** too - Cluely, and Otter on its Enterprise tier. Raven does both jobs in one open-source, local-first app that brings your own API keys, keeps recordings on your machine, and never puts a bot in the meeting.
Screen-share invisibility, free - vs Cluely's \$149.99/mo tier. Local vs cloud, open vs closed.
Unlimited AI notes on your own key - vs Fathom's 5-summaries-a-month free cap. Plus live help during the call.
Live assist and unlimited local history - vs Granola's 30-day free window and no in-call help.
No 300-minute cap, no bot, no cloud - vs Otter's metered free tier.
## The short version
| | Raven | Cluely | Fathom | Granola | Otter |
| ------------------------- | ------------------------ | ------------------- | --------------------------- | ------------------- | --------------- |
| Live in-call assist | Yes | Yes | Live notes only | No | Enterprise only |
| Invisible to screen share | Yes, free | Only on \$149.99/mo | No | No | No |
| Bot-free capture | Yes (Mac + Windows) | Yes | Bot, or bot-free beta (Mac) | Yes (Mac + Windows) | Bot |
| Where your data lives | Local SQLite | Cloud | Cloud | Cloud | Cloud |
| Open source | Yes | No | No | No | No |
| Price | Free (your own API keys) | \$19.99-\$149.99/mo | \$16/user/mo+ | \$14/user/mo | \$8.33/user/mo+ |
Raven is free and open source. You bring your own keys, so transcription (Deepgram / AssemblyAI) and AI (Anthropic / OpenAI) are billed by those providers for what you use - typically a few dollars a month for regular use, and nothing when you're not in a call.
Competitor prices verified August 2026 - they change often; check each vendor's pricing page before relying on a number.
# Zoom AI Companion, Gemini, Teams Copilot - do you still need a meeting assistant?
Source: https://docs.useraven.ai/compare/platform-built-ins
How Raven compares to platform built-ins like Zoom AI Companion, Gemini in Google Meet, and Microsoft Teams Copilot - one local tool across every platform, with live help and no vendor lock-in.
"Isn't this already included in Zoom / Google / Microsoft?" is a fair question - the platforms now ship their own AI: **Zoom AI Companion**, **Gemini in Google Meet**, and **Microsoft Teams Copilot**. If you live entirely inside one platform and only want a recap, they're genuinely fine and usually cost nothing extra.
The catch is that they are **built for their own platform, on their own cloud** - and they don't help you *during* the call. Here's the honest comparison.
## At a glance
| | Raven | Platform built-ins (Zoom / Meet / Teams) |
| ------------------------------------------------ | ---------------------------------------------- | ------------------------------------------------------------------------------------------------- |
| Works across Zoom, Meet, Teams **and** in-person | **Yes** - captures any call plus in-room audio | No - each one only works inside its own platform |
| Live in-call assist ("what should I say?") | **Yes** - hidden overlay during the call | No - summaries arrive after the meeting |
| Invisible to screen share | **Yes** | n/a - no in-call overlay |
| Where your data goes | **Local SQLite**; audio uses **your** API keys | The platform vendor's cloud, under your org's tenant and terms |
| One unified history across all your meetings | **Yes** - one local Ask index | No - memory is split across three vendors |
| Cost | **Free** (your own API usage) | Bundled on some plans; Teams Copilot typically needs a paid add-on (M365 Copilot, \~\$30/user/mo) |
| Open source | **Yes** - auditable | No |
## Three real limits of the built-ins
1. **They're platform-locked.** Zoom AI Companion only helps in Zoom, Gemini only in Meet, Copilot only in Teams. If your week spans all three - plus a phone call or an in-person meeting - your notes and history are fragmented across vendors, with different formats and no single search. Raven is **one tool** for every call and in-room audio.
2. **They're after-the-fact.** They transcribe and summarize once the meeting ends. None of them sit beside you *during* the call suggesting what to say, surfacing an answer to a hard question, or running a live sales/interview checklist - and none are hidden from screen share. That live moment is Raven's whole reason to exist.
3. **Your meetings feed the platform vendor.** Content is processed in Microsoft's, Google's, or Zoom's cloud under your organization's agreement and admin controls (which can also switch the feature off). Raven keeps transcripts and notes **on your machine** and only sends audio to the transcription/AI providers **you** chose with **your own** keys - and because it's open source, that data path is auditable.
## When the built-ins are the right call (honestly)
* You use **exactly one** platform and never leave it.
* You only want a **post-meeting recap**, not live help.
* It's **already included** in your plan and you don't want to install anything.
If that's you, use what's bundled - it's zero-setup and free on most Zoom and Google plans.
## When Raven fits better
* You bounce between **Zoom, Meet, Teams, phone, and in-person** and want **one** consistent tool and history.
* You want **help during** the call, invisible on screen share.
* You want meetings kept **local** and a tool you can **audit** - not another feed into a platform vendor.
* You'd rather pay your AI provider **cents** than add a per-seat Copilot license.
Free, open source, Mac + Windows. Works across every platform.
How Raven stacks up against Fathom, Granola, Otter, and more.
Platform built-ins change fast and their availability depends on your plan and admin settings (e.g., Teams Copilot generally requires a Microsoft 365 Copilot add-on). Check what your organization actually has enabled.
# Raven vs Read.ai
Source: https://docs.useraven.ai/compare/read-ai
A free, open-source Read.ai alternative - local-first, no bot, no per-seat fee, with live in-call help instead of after-the-fact analytics.
Read.ai leans into meeting **analytics and coaching** - engagement scores, sentiment, "meeting coach" readouts - captured by a **bot** that joins your calls and stored in its **cloud**, per seat. Raven skips the bot and the cloud: capture is **local**, it's **free** and **open source**, and it helps you **live during** the call.
## At a glance
| | Raven | Read.ai |
| ------------------------ | --------------------------------- | ---------------------------------------------------------------------------- |
| Live in-call assist | **Yes** | No - analytics after the call |
| Capture | Local, bot-free (Mac + Windows) | Bot joins Zoom/Meet/Teams |
| Where your meetings live | Local SQLite | Read.ai's cloud (lost if you cancel) |
| Price | Free, open source (your own keys) | Free (5 meetings/mo); Pro \$19.75/user/mo (\$15 annual); Enterprise \$29.75+ |
| Open source | Yes | No |
## Price
Read.ai's free plan caps you at **5 meetings/month**; unlimited transcripts start at **Pro \$19.75/user/mo** (\$15 annual), and video playback is gated to **Enterprise \$29.75+/user/mo**. Recordings are only accessible while you keep paying.
Raven is free; you pay your own AI/transcription usage (a few dollars a month for typical use), keep everything locally, and nothing disappears if you stop.
## What Read.ai does better (honestly)
* Meeting **analytics/coaching** dashboards, engagement metrics, and cross-workspace search.
* Mobile apps and a polished hosted experience.
If you want engagement analytics across a team's meetings, Read.ai is built for that.
## When Raven wins
* You want **help during** the call, not scorecards after.
* You want meetings **local** and permanent - not a cloud archive you lose on cancel.
* You want **no bot** in the call and a **free**, auditable tool.
Free, open source, Mac + Windows. Bring your own keys.
Raven vs the whole category, honestly ranked.
Read.ai pricing verified August 2026 (read.ai/plans-pricing). Vendors change pricing often - confirm before relying on a figure.
# Raven vs SalesEQ
Source: https://docs.useraven.ai/compare/saleseq
A free, open-source, voice-only and local alternative to SalesEQ - no buyer video analysis, no cloud, and live help during the call.
SalesEQ analyzes the **nonverbal** side of a sales call - tone, hesitation, confidence, **facial expressions, and body language** - by recording **video** (via Recall.ai's desktop SDK) and analyzing it in the **cloud**, wired into your CRM. Raven takes the opposite, deliberate approach: **voice-only, on your machine, no camera**, open source and free - and it helps you **live during** the call rather than scoring it after.
## At a glance
| | Raven | SalesEQ |
| ----------------------------- | --------------------------------- | ------------------------------------------------------------- |
| What it analyzes | Words + voice (You / Them) | Words + voice + **video: facial expressions & body language** |
| Camera / buyer video required | **No** - audio only | Yes - captures the video call |
| Capture | Local, on-device (Mac + Windows) | Recall.ai Desktop SDK; processed in SalesEQ's cloud |
| Where your data lives | Local SQLite | SalesEQ's cloud + CRM |
| Live in-call assist | **Yes** | No - analysis after the call |
| Price | Free, open source (your own keys) | Not public - sales-team product (contact sales) |
| Open source | Yes | No |
## The real difference: voice-only and local, by design
SalesEQ's pitch is reading the buyer's **face and body language** on video. Raven deliberately **doesn't** do that. Recording and scoring the other person's facial expressions is privacy-invasive, consent-heavy, and scientifically contested - so Raven focuses on **what was said and how it sounded**, split into You/Them streams, analyzed **on your machine**. You get talk ratio and grounded sentiment without pointing a model at your buyer's face.
And where SalesEQ uploads audio **and video** to its cloud, Raven keeps transcripts and notes in **local SQLite**; only audio goes to the transcription/AI providers you chose with your own keys.
## What SalesEQ does better (honestly)
* **Nonverbal / video intelligence** - if you believe facial and body-language signals help you sell, SalesEQ is purpose-built for exactly that, with reliable Recall-powered capture and CRM push.
If video-based buyer analysis is your thing, Raven is not that tool and isn't trying to be.
## When Raven wins
* You'd rather analyze **voice + words**, not put a model on your buyer's **face**.
* You want capture and storage kept **local**, and a tool you can **audit** (open source).
* You want **live help during** the call, and it to be **free** - no cloud, no CRM contract.
Free, open source, Mac + Windows. Bring your own keys.
You-vs-Them share, action items, follow-up email.
SalesEQ details verified August 2026 (saleseq.ai; Recall.ai case study). SalesEQ does not publish pricing. Confirm current details with the vendor.
# Raven vs Sembly
Source: https://docs.useraven.ai/compare/sembly
A free, open-source Sembly alternative - live in-call help and local notes, with no per-seat cloud subscription (Sembly retired its free plan in 2026).
Sembly is a cloud notetaker with structured minutes, task detection, and analytics, captured by a **bot**. As of 2026 it **retired its free plan** (now a 7-day trial), and it's **per seat**. Raven keeps a **free** tier, runs **local** with **no bot**, is **open source**, and helps you **live during** the call.
## At a glance
| | Raven | Sembly |
| ------------------------ | --------------------------------- | ---------------------------------------------------------- |
| Live in-call assist | **Yes** | No - minutes after the call |
| Free tier | **Yes** - full app | **No** (retired March 2026; 7-day trial) |
| Capture | Local, bot-free (Mac + Windows) | Bot joins the call |
| Where your meetings live | Local SQLite | Sembly's cloud |
| Price | Free, open source (your own keys) | Basic \$10/mo; Pro \$20/user/mo; MAX \$30/user/mo (annual) |
| Open source | Yes | No |
## Price
Sembly's old free Personal plan is gone. Paid plans (annual) are **Basic \$10/mo** (1 user), **Pro \$20/user/mo**, and **MAX \$30/user/mo** (3-seat minimum), all higher month-to-month.
Raven is free; you pay your own AI/transcription usage (a few dollars a month for typical use), no seat fee.
## What Sembly does better (honestly)
* Structured **meeting minutes**, automatic task/risk detection, and an analytics dashboard.
* Broad integrations and multi-meeting AI chat on higher tiers.
If you want automated formal minutes across a team, Sembly is designed for that.
## When Raven wins
* You want a real **free** tier - not a 7-day trial then per-seat billing.
* You want **live help during** the call, meetings kept **local**, and **no bot**.
* You want a **free**, **open-source**, auditable tool.
Free, open source, Mac + Windows. Bring your own keys.
Raven vs the whole category, honestly ranked.
Sembly pricing verified August 2026 (sembly.ai/pricing). Vendors change pricing often - confirm before relying on a figure.
# Raven vs Tactiq
Source: https://docs.useraven.ai/compare/tactiq
A free, open-source Tactiq alternative - native local capture and live in-call help, not a browser-only transcript extension with AI-credit caps.
Tactiq is a lightweight **Chrome extension** that shows live transcripts in Google Meet/Zoom/Teams and runs AI on a **credit** system. It's cheap and handy, but it's **browser-only** (no native desktop or in-person capture), it doesn't help you decide **what to say**, and its AI is **metered by credits**. Raven captures **natively and locally**, adds **live assist**, and is **free** and **open source** with unlimited AI on your own key.
## At a glance
| | Raven | Tactiq |
| ------------------------ | -------------------------------------------------- | ------------------------------------------------------- |
| Capture | Native local (Mac + Windows), any call + in-person | Chrome extension only (Meet/Zoom/Teams tabs) |
| Live in-call assist | **Yes** ("what should I say?") | Live transcript only |
| AI limits | Unlimited (your own key) | Free 5 credits/mo; Pro 10 credits/mo |
| Where your meetings live | Local SQLite | Tactiq's cloud |
| Price | Free, open source (your own keys) | Free (10 transcripts/mo); Pro \$8/user/mo; Team \$16.67 |
| Open source | Yes | No |
## Price
Tactiq's free plan gives **10 transcripts + 5 AI credits/month**. **Pro (\$8/user/mo annual, \$12 monthly)** makes transcripts unlimited but still caps AI at **10 credits/month** - unlimited AI needs **Team \$16.67/user/mo**.
Raven's AI is unlimited on your own key (a few dollars a month of usage), with no credits and no seat fee.
## What Tactiq does better (honestly)
* **Zero install** - it's a browser extension, great for Chromebooks and quick live transcripts.
* Very cheap entry point and a clean share/export flow.
If you just want a live transcript in your browser and rarely run AI, Tactiq is a tidy, inexpensive option.
## When Raven wins
* You want **native** capture (any call, in-person) - not browser tabs only.
* You want **live assist**, not just a transcript, and **unlimited AI** without credits.
* You want meetings **local** and a **free**, open-source, auditable tool.
Free, open source, Mac + Windows. Bring your own keys.
Raven vs the whole category, honestly ranked.
Tactiq pricing verified August 2026 (tactiq.io/buy). Vendors change pricing often - confirm before relying on a figure.
# Raven vs tl;dv
Source: https://docs.useraven.ai/compare/tldv
A free, open-source tl;dv alternative with live in-call help and talk ratio - no per-seat pricing and no jump to a \$59 sales tier. Meetings stay local.
tl;dv is a solid sales-coaching notetaker: it records your calls and, on its **Business** tier, scores reps with MEDDIC/BANT/SPIN playbooks. The cost of that is the price curve - Pro is **\$18/user/mo** and Business jumps to **\$59/user/mo** - and it's cloud, per-seat, and after-the-call. Raven helps you **live during** the call, shows **talk ratio** on every session, keeps meetings **local**, and is **free** and **open source**.
## At a glance
| | Raven | tl;dv |
| ------------------------ | ---------------------------------- | -------------------------------------------------------------------- |
| Live in-call assist | **Yes** - sales modes, suggestions | No - coaching is after the call |
| Talk ratio (You vs Them) | **Yes**, on every session | On paid tiers |
| Where your meetings live | Local SQLite | tl;dv's cloud (EU-hosted) |
| Price | Free, open source (your own keys) | Free (10 AI notes); Pro \$18/user/mo; Business \$59/user/mo (annual) |
| Capture | Local, bot-free (Mac + Windows) | Bot or bot-free modes |
| Open source | Yes | No |
## Price
tl;dv's free plan records unlimited calls but caps you to **10 AI notes** and auto-deletes after 3 months. Real use starts at **Pro \$18/user/mo** (annual; \$29 monthly), and the sales-coaching features - scorecards, objection analysis, playbooks - live in **Business at \$59/user/mo** (annual; \$98 monthly). For a 5-rep team that's roughly \$3,540/year on Business.
Raven is free; you pay your own AI/transcription usage (a few dollars a month for typical use), with no per-seat fee and no coaching-tier upsell.
## Where your meetings live
tl;dv stores recordings in its (EU-hosted) cloud. Raven keeps transcripts and notes in a **local SQLite database** on your machine, splits **You** and **Them** on-device, and only sends audio to the providers you chose with your own keys. Open source means the data path is auditable.
## What tl;dv does better (honestly)
* **Sales coaching**: automated MEDDIC/BANT/SPIN scorecards and objection-handling analysis on Business.
* Native **CRM** sync and manager review workflows.
* EU hosting and a polished team experience.
If you're a sales manager coaching a floor of reps and want after-call scorecards, tl;dv's Business tier is purpose-built.
## When Raven wins
* You want **live** help in the call (next question, objection response) - not just an after-call score.
* You want **talk ratio** and sales **modes** without paying for a \$59/seat tier.
* You want meetings **local** and a tool you can **audit**.
* You'd rather pay your AI provider **cents** than per seat.
Raven ships live sales-mode assist and talk ratio today; automated call-scoring checklists are on the roadmap, not shipped yet. If after-call scorecards are your core need, weigh that honestly.
## Switching from tl;dv
1. Export anything you want to keep (tl;dv free auto-deletes after 3 months).
2. [Download Raven](https://useraven.ai/download), add your keys, pick a sales Mode.
3. Run both for a week; cancel tl;dv once Raven covers your daily habit.
Free, open source, Mac + Windows. Bring your own keys.
How Raven helps during the call.
tl;dv pricing verified August 2026 (tldv.io). Vendors change pricing often - confirm before relying on a figure.
# Architecture
Source: https://docs.useraven.ai/contributing/architecture
How Raven's processes, natives, RAG, and session context fit together.
Raven is an **Electron** app. There is no Raven server in this build. The diagram below covers both the real-time capture path and the post-call layer (Ask, retrieval, recap, auto-start).
## Layers
| Layer | Role |
| ------------------------------- | ------------------------------------------------------- |
| **Renderer** (React + Tailwind) | Overlay and dashboard |
| **Preload** | `window.raven` context bridge |
| **Main** | Capture orchestration, STT, Assist, SQLite, settings |
| **Native** | WASAPI / ScreenCaptureKit capture + GStreamer AEC addon |
## Data flow
```mermaid theme={null}
flowchart TB
subgraph capture [Native capture]
Win["Windows: WASAPI loopback + capture"]
Mac["macOS: ScreenCaptureKit + CoreAudio"]
end
subgraph local [Real-time on this machine]
SAN[systemAudioNative]
AEC["GStreamer webrtcechoprobe / webrtcdsp"]
REG[ResidualEchoGate]
AM[audioManager]
RAG["ragService - MiniLM"]
MEM["sessionMemory - this recording"]
CS[claudeService]
DB[("SQLite data/raven.db")]
CFG[electron-store]
OV[Overlay]
Dash[Dashboard]
end
subgraph postcall [Post-call on this machine]
SIDX["sessionIndexService - MiniLM transcript index"]
SQA["sessionQaService - Ask, streamed"]
RECAP["Recap - action items, follow-up email, export, insights"]
end
subgraph byok [Your API keys]
STT["Deepgram nova-3 and/or AssemblyAI u3-rt-pro"]
LLM[Anthropic or OpenAI]
end
Win --> SAN
Mac --> SAN
SAN -->|"system PCM = Them"| AM
SAN --> AEC
AEC -->|"cleaned mic"| REG
REG -->|"You if not echo"| AM
AM --> STT
STT --> OV
OV --> CS
MEM --> CS
RAG --> CS
CS --> LLM
CS --> DB
RAG --> DB
Dash --- DB
MD["meetingDetector - window titles"] -.->|"offer to start"| AM
DB -->|"saved transcript"| SIDX
SIDX -->|"chunks + embeddings"| DB
SQA -->|"retrieve top chunks"| DB
SQA --> LLM
RECAP --> DB
RECAP --> LLM
Dash --- SQA
Dash --- RECAP
```
1. Native capture grabs system + mic (16 kHz PCM).
2. System PCM is the AEC **reference**. Mic goes through GStreamer AEC3, then `ResidualEchoGate`.
3. Two STT connections: **You** (cleaned mic) and **Them** (system).
4. Overlay shows the live transcript. Assist uses mode brief + RAG + in-RAM session memory + transcript tail.
5. Non-incognito sessions persist in SQLite. Settings and keys live in encrypted `electron-store`.
The diagram above is the real-time capture + Assist path. A separate **post-call layer** runs on the saved transcript: structured [action items, a follow-up email draft, talk ratio, and export](/features/meeting-intelligence), and [Ask](/features/ask) (Q\&A over one session or your whole history, using an on-device MiniLM transcript index). Retrieval and embeddings stay local; only your own model call goes out. [Meeting auto-start](/features/meeting-detection) watches window titles to offer recording - no bot joins.
## What “memory” means
| Mechanism | Lifetime | Used for |
| ----------------------------------- | --------------------------------------- | ----------------------------------------------------------------------- |
| Live Assist turns + `sessionMemory` | Current recording, RAM | Running summary, pinned opening, pinned questions, last turns |
| SQLite sessions / messages | Until you delete them | History, summaries, insights, chat replay |
| RAG chunks | Until you remove the file from the mode | Retrieve-then-prompt on Assist |
| Session Ask index | Until you delete the session | Retrieve-then-prompt for **Ask across all meetings** (on-device MiniLM) |
| Ask conversations | Until you delete the chat / session | Persisted Ask history (per-session + standalone threads) |
| electron-store | Until you reset settings | Keys and preferences - not semantic memory |
There is no global “Raven remembers you across meetings” store.
## Key modules
| Module | File | Role |
| ------------------ | ----------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------- |
| Audio manager | `src/main/audioManager.ts` | Recording lifecycle, STT engine pick |
| Native capture | `src/main/systemAudioNative.ts` | Platform capture + AEC wiring |
| Residual echo gate | `src/main/residualEchoGate.ts` | Drop leftover speaker bleed before You STT |
| Deepgram | `src/main/transcriptionService.ts` | Dual WebSocket `nova-3` |
| AssemblyAI | `src/main/services/assemblyAITranscriptionService.ts` | Dual-stream `u3-rt-pro` |
| STT routing | `src/shared/sttCapabilities.ts` | Language + preference |
| Assist | `src/main/claudeService.ts` | Overlay AI (Claude / OpenAI) |
| Session memory | `src/main/services/ai/sessionMemory.ts` | In-recording compression |
| RAG | `src/main/services/ragService.ts` | Parse, embed, retrieve |
| SQLite | `src/main/services/database.ts` | Sessions, modes, RAG chunks, Ask index + chats |
| Settings | `src/main/store.ts` | Encrypted electron-store |
| Sessions | `src/main/services/sessionManager.ts` | Autosave, incognito |
| Ask + retrieval | `services/sessionQaService.ts`, `services/sessionIndexService.ts` | Streamed Q\&A over one session or your whole history (MiniLM index) |
| Post-call recap | `services/followupEmailService.ts`, `services/sessionExportService.ts`, `services/insightsService.ts` | Action items, follow-up email, export, insights |
| Meeting detect | `src/main/meetingDetector.ts` | Window-title auto-start - no bot |
## Windows
| Window | Purpose |
| ------------- | --------------------------------------------- |
| **Overlay** | Transcript, Assist, pill, stealth / incognito |
| **Dashboard** | History, settings, modes, insights, Ask |
## Natives
| Platform | Capture | Echo cancellation |
| -------- | --------------------------------------------------- | ------------------------------------------ |
| macOS | Swift `audiocapture` (ScreenCaptureKit + CoreAudio) | `raven-aec.node` (GStreamer + WebRTC AEC3) |
| Windows | Rust/NAPI `raven-windows-audio` (WASAPI) | Same C++ addon |
Not in this product: hosted Pro, login, cloud sync, Recall meeting-bot capture.
# Contributing
Source: https://docs.useraven.ai/contributing/repo-structure
How to contribute to the Raven open-source project.
Raven is open-source and we welcome contributions. Here's how to get involved.
## Getting the Code
```bash theme={null}
git clone https://github.com/Laxcorp-Research/project-raven.git
cd project-raven
```
For the full build-from-source guide including native dependencies, see:
* [macOS Setup](/quickstart/macos)
* [Windows Setup](/quickstart/windows)
## Making Changes
1. Fork the repo on GitHub
2. Create a branch for your change
3. Make your changes and test them
4. Submit a pull request
### Commit guidelines
* One feature or fix per commit
* Use clear commit messages: `feat: add X`, `fix: resolve Y`
* Run `npm test` and `tsc --noEmit` before pushing
### What not to commit
* `.env` files, API keys, or any credentials
* Generated files (build artifacts, `node_modules`)
## CI Checks
Every pull request automatically runs:
* **TypeScript type checking** - `tsc --noEmit`
* **Unit and integration tests** - Vitest
Both must pass before a PR can be merged.
## Where to Start
* **Issues** - check the [GitHub Issues](https://github.com/Laxcorp-Research/project-raven/issues) for bugs and feature requests
* **Discussions** - ask questions or propose ideas in [GitHub Discussions](https://github.com/Laxcorp-Research/project-raven/discussions)
## License
Raven is released under the [MIT License](https://github.com/Laxcorp-Research/project-raven/blob/main/LICENSE).
# Testing Guide
Source: https://docs.useraven.ai/contributing/testing-guide
How to test Raven - automated tests and manual QA checklist.
## Automated Tests
```bash theme={null}
npm test # Unit + integration tests (Vitest)
npm run test:coverage # With coverage report
npm run test:e2e # End-to-end tests (requires `npm run build` first)
npm run test:all # Everything
```
## Manual Testing Checklist
Before a release or after significant changes, run through these scenarios.
### First Launch & Onboarding
1. Delete app data: `~/Library/Application Support/Raven` or `project-raven` (macOS); `%APPDATA%\Raven` or `project-raven` (Windows)
2. Run `npm run dev`
3. Verify: onboarding appears, you can enter API keys, overlay appears after completing setup, tray icon shows in the menu bar
### Recording & Transcription
1. Join a call or play a video with speech
2. Press `Cmd+Shift+Space` (`Ctrl+Shift+Space` on Windows) to start recording
3. Talk for 2+ minutes, let the other side talk too
**Verify:**
* Your speech shows as "You" (right side), theirs as "Them" (left side)
* Interim text appears while speaking, replaced by final text on pause
* No echo - your voice doesn't appear on the "Them" side
* Tray icon shows recording state
### AI Assistance
1. While recording, click **Assist** or press `Cmd+Enter`
2. Try "What should I say?", "Follow-up", "Recap"
3. Type a specific question in the input bar
4. Stop recording, then press `Cmd+Enter` on a webpage
**Verify:**
* Responses reference the actual conversation
* "What should I say?" references the last thing the other person said
* Typed questions get direct answers
* Screenshot mode (no session) describes and solves what's on screen
### Overlay & Stealth
1. Position the overlay over a meeting window
2. Share your screen - ask someone if they can see the overlay
3. Drag the overlay by the pill, resize from edges
4. Click through transparent areas to the app behind
**Verify:**
* Overlay is invisible to screen sharing
* Drag and resize work smoothly
* Clicks pass through transparent areas
### Keyboard Shortcuts
| Shortcut | Expected |
| ------------------------- | ------------------------- |
| `Cmd + \` | Toggle overlay visibility |
| `Cmd + Shift + Space` | Start/stop recording |
| `Cmd + Enter` | AI assist with screenshot |
| `Cmd + Shift + Backspace` | Clear conversation |
### Session Persistence
1. Record a 2+ minute conversation, get an AI response
2. Stop recording and quit the app (`Cmd+Q`)
3. Reopen the app
**Verify:**
* Session appears in the dashboard
* Title was auto-generated
* Full transcript and AI responses are preserved
### Graceful Shutdown
1. Start recording, then quit (`Cmd+Q`) while still recording
**Verify:**
* App quits within a few seconds (no hang)
* No orphan processes in Activity Monitor
* The interrupted session is saved and visible on reopen
### Modes
1. Switch to "Interview" mode in the dashboard
2. Start recording and ask for AI help
**Verify:**
* AI responses reflect the mode's personality
* Can create and edit custom modes
* Can upload documents and see them listed
### Edge Cases
* **No internet during recording**: disconnect WiFi, reconnect after 10 seconds - transcript should pause and resume
* **Rapid start/stop**: start and stop recording within 2 seconds - no crashes
* **Very long utterance**: speak for 60+ seconds without pausing - text keeps updating
# AI Assistance
Source: https://docs.useraven.ai/features/ai
Overlay Assist with your Anthropic or OpenAI key, session context, and optional screenshots.
Raven calls **your** Anthropic or OpenAI key from the desktop app. There is no Raven AI proxy.
## Providers and models
Configure **Settings → API Keys** (and the model / effort pickers).
Examples from the in-app catalog (the list in Settings is the source of truth):
| Provider | Faster | More capable |
| ------------- | -------------------------- | ------------------------------------------------ |
| **Anthropic** | Claude Haiku 4.5 | Claude Sonnet 5 / 4.6, Claude Opus 5 / 4.8 / 4.7 |
| **OpenAI** | GPT-5.6 Luna, GPT-5.4 Mini | GPT-5.6 Terra / Sol, GPT-5.5, GPT-5.4 |
Haiku or Luna are a good default for live Assist. Pick effort (low → max) when the model supports it.
## Overlay actions
| Action | What it does |
| ---------------------- | --------------------------------------------- |
| **Assist** | Help from the live moment and your mode brief |
| **What should I say?** | Suggested reply to the last **Them** turn |
| **Follow-up** | A follow-up question |
| **Recap** | Summary so far |
| **Tell me more** | Expand the last card |
`Cmd/Ctrl + Enter` runs Assist and includes a screenshot when requested.
You can also type a question in the overlay input. Mode instructions (topics, tone, canned replies) are treated as **instructions**, not optional flavor.
## Session context (not long-term memory)
Each Assist call is a new API request. For a long recording Raven sends:
1. A compressed **session memory** (cheap model, in RAM)
2. Pinned **opening transcript** and your typed questions
3. The last few turns
4. The live transcript tail
That memory dies when the session ends. There is no cross-meeting user profile. History you reopen later lives in **SQLite**, not in that RAM structure.
## RAG
If the active **mode** has documents, Assist retrieves the top MiniLM matches and injects them into the system prompt. See [Modes](/features/modes).
## Screenshots
`Cmd/Ctrl + Enter` can attach a PNG from `desktopCapturer`. Raven tries to keep its own overlay out of that frame.
* **During a recording:** screenshot + transcript + mode + RAG
* **Not recording:** visual content only
## Formatting
Responses render markdown, KaTeX math, and highlighted code.
## Copy
Hover a response card and use the copy control to put it on the clipboard.
# Ask
Source: https://docs.useraven.ai/features/ask
Chat with a single meeting or across your whole history - local retrieval, your own model, streamed answers.
**Ask** turns your transcripts into a conversation. It runs locally: retrieval and embeddings are on-device, and only your own model call (the Anthropic or OpenAI key you set for notes) leaves your machine. Answers stream in token-by-token.
## Two ways to ask
The **Ask** tab on any session answers from that call's transcript. Good for "how did I do?", "what did I commit to?", or any detail from the conversation.
**Ask across all meetings** (from the dashboard header) searches your whole history and cites the sessions it drew from. Click a source to open it.
## A thinking partner, not a search box
Ask reads the transcript and talks it through with you - sentiment, what landed, what you may have missed - and answers direct factual questions from what was actually said. It won't invent names, numbers, or dates that aren't in the transcript; if something isn't there, it says so.
Follow-ups build on earlier turns. Long chats keep their context: older turns are folded into a running summary (never dropped) so the thread stays coherent without exceeding the model's context window.
## Saved conversations
* **Per meeting:** each session keeps a single Ask chat, restored when you reopen it.
* **Across all meetings:** keep as many chats as you want - **New chat**, switch, **rename**, or **delete** from the sidebar. Each is auto-titled from your first question and saved locally.
Answers stream as they generate, and the input stays pinned to the bottom like a normal chat. If you navigate away mid-answer, the turn isn't lost - generation finishes in the background and the full answer is saved.
## How retrieval works (across all meetings)
Every session's transcript is chunked and embedded locally with **`Xenova/all-MiniLM-L6-v2`**. New sessions are indexed automatically; older ones are backfilled the first time you open Ask.
Your question is embedded and ranked against the chunks by cosine similarity.
The top matches (within a token budget) plus recent conversation go to your model, which answers with inline citations like `[1]`.
The per-meeting Ask tab skips retrieval - a single call's transcript usually fits, so it's fed directly for the most accurate read.
## Privacy
Chunks, embeddings, and saved chats live in **SQLite** on your machine. There is no Raven server and no cross-account sync. Deleting a session also removes its transcript chunks and its Ask chat.
# Meeting auto-start
Source: https://docs.useraven.ai/features/meeting-detection
Raven can notice a Zoom, Meet, Teams, or Webex meeting and offer to start - no bot ever joins.
Raven can watch for a meeting window and start recording so you don't forget to hit record. **No bot joins your call** - detection just reads the titles of open windows on your own machine, and capture stays local as always.
## Modes
Set this in **Settings → General**:
| Mode | Behavior |
| -------------------- | ------------------------------------------------------------------ |
| **Off** | No detection. Raven does no window scanning at all. |
| **Prompt** (default) | Shows a toast when a meeting appears; you decide whether to start. |
| **Auto** | Starts a session automatically when a meeting is detected. |
## What it detects
| Platform | Detected | Ignored |
| ------------------- | ------------------------------------- | ------------------------------------ |
| **Zoom** | In-meeting / webinar window | The idle Zoom app / Workplace |
| **Google Meet** | An in-call tab (matched by room code) | The `meet.google.com` landing page |
| **Microsoft Teams** | Scheduled meetings **and 1:1 calls** | Chat, Calls, Calendar, Activity tabs |
| **Cisco Webex** | Meeting window | The main Webex app |
Raven notifies **once** per meeting and won't nag again until that meeting window disappears. It's also disabled automatically while you're already recording.
On macOS, reading window titles uses ScreenCaptureKit, so auto-start needs the same **Screen Recording** permission that recording uses. Without it, nothing is detected.
# Post-call recap
Source: https://docs.useraven.ai/features/meeting-intelligence
Action items, a follow-up email draft, talk ratio, and Markdown/PDF export - all from your own key.
When a call ends, Raven turns the transcript into things you can act on. Everything uses the model key you set in **Settings → API Keys**; nothing goes to a Raven server.
## Action items
The **Summary** tab shows a structured checklist pulled from the transcript - the task, who owns it, and a deadline when one was spoken. Tick items off as you go; the checked state is saved with the session.
## Follow-up email
**Draft follow-up email** turns the call into a ready-to-send recap - greeting, what was discussed, action items, and a sign-off with your name. It costs a model call, so the draft is **saved with the session** and is still there when you come back. Regenerate or copy it any time.
## Talk ratio
Your **You vs. Them** speaking share, by word count (approximate). It appears live in the overlay's transcript during a call, on each session afterward, and in exports - a quick gut check on whether you dominated or listened.
## Export
Save any session recap with the **Export** menu:
| Format | Contents |
| -------------------- | -------------------------------------------------------- |
| **Markdown** (`.md`) | Title, date, duration, talk ratio, action items, summary |
| **PDF** (`.pdf`) | The same recap, print-styled |
Tick **Include transcript** to append the full You / Them transcript. You choose where the file is saved.
## Insights
The **Insights** tab runs deeper analysis on demand - sentiment, topics, and key phrases - using your key. See [Sessions](/features/sessions).
These recap tools generate automatically after a call (action items) or on one click (email, export, insights). They never run against a hosted service - the transcript stays local and only your model call goes out.
# Modes
Source: https://docs.useraven.ai/features/modes
Local AI profiles, quick actions, notes templates, and on-device RAG.
A **mode** is a local behavior profile: system prompt, optional quick actions, notes template, and optional documents. Templates are bundled in the app. There is no server prompt fetch.
## Built-in templates
Raven seeds **General Assistant**. You can create modes from templates such as:
| Template | Typical use |
| -------------------- | ------------------------ |
| **Interview** | STAR-style answers |
| **Sales Call** | Objections and close |
| **Meeting Notes** | Decisions and actions |
| **Job Search** | Screens and offer talks |
| **Learning** | Lectures and study |
| **Coding Interview** | Timed coding assessments |
## Create or edit
**Dashboard → Modes**:
1. Start from a template or blank
2. Write the system prompt (this is **instructions**, not optional flavor)
3. Add overlay quick actions
4. Define notes sections used when summarizing
5. Upload reference documents for RAG
## Quick actions
Custom overlay buttons. Each sends a fixed prompt plus the live transcript and mode brief.
## Document upload (RAG)
1. Add `.pdf`, `.docx`, `.txt`, or `.md` (max **10 MB** each) in the mode editor
2. The file is parsed, chunked (\~500 words, overlap), and embedded on-device with `Xenova/all-MiniLM-L6-v2` (first run may download \~30 MB)
3. Vectors stay in SQLite (`mode_context_chunks`)
4. On Assist, Raven retrieves the top matches (token-capped) and injects them into the system prompt
Nothing is uploaded to a Raven server. The embedding model and chunks stay on disk.
### What to upload
* Interview: JD, resume, company notes
* Sales: pricing, product specs
* Meetings: briefs, prior notes
### Limits
* Documents belong to **that mode** only
* Re-upload replaces the previous file
* Local files only - no URL ingest
* Retrieval is cosine similarity in-process, not a hosted vector DB
## Switching
The active mode applies to new sessions. Each saved session records which mode was used.
## Reset
Built-in templates can be reset to the bundled prompt. Custom modes can be edited or deleted. Modes do not sync across machines.
# Overlay & Stealth
Source: https://docs.useraven.ai/features/overlay
Floating overlay, content protection, tray, and click-through.
The overlay is a transparent panel for transcript, Assist, and the controller pill. Stealth uses Electron `setContentProtection` so typical screen-share APIs omit the overlay and dashboard. That is not a guarantee against every capture tool.
## Stealth
On by default.
* Zoom, Meet, Teams, Discord usually cannot see it
* You still see and use the overlay
| Where | How |
| ---------------- | -------- |
| Overlay pill | Eye icon |
| Dashboard header | Eye icon |
| Settings | General |
`Cmd/Ctrl + \` hides or shows the overlay **on your** screen. That is separate from stealth.
### Platform notes
* **macOS**: Content protection plus dock hide when stealth is on. Tray can hide via the same path.
* **Windows**: Content protection. A new tray icon may sit in the `^` overflow until you pin it.
* **Linux**: Not supported.
* **OBS**: Display capture usually respects protection. Game / window capture sometimes does not.
## Controller pill
| Control | What it does |
| ------------- | ------------------------------------------ |
| **Logo** | Expand / compact the overlay |
| **Record** | Start or stop (`Cmd/Ctrl + Shift + Space`) |
| **Stealth** | Screen-share invisibility |
| **Incognito** | Do not persist this session to SQLite |
| **Hide** | Hide overlay (`Cmd/Ctrl + \` to restore) |
There is no Fast/Deep toggle. Pick model and effort in Settings.
## Tabs
| Tab | Content |
| -------------- | ------------------------------- |
| **Responses** | Streaming Assist cards |
| **Transcript** | You / Them |
| **Chat** | Overlay thread for that session |
## Move and resize
* Drag the pill
* Resize from the edges
* Double-click a resize edge to reset
On Windows, `Ctrl + \` re-shows a hidden overlay and keeps it grabbable.
## Click-through
Transparent pixels pass clicks to the app beneath. Buttons, cards, and resize handles capture clicks.
## Tray
* Start / stop listening
* View sessions
* Settings
* Quit
Left-click on Windows/Linux reopens the dashboard. Packaged builds load icons from `resources/tray`.
# Recording & Transcription
Source: https://docs.useraven.ai/features/recording
Dual-stream capture, GStreamer AEC, residual echo gate, and Deepgram / AssemblyAI routing.
## Starting a recording
Press `Cmd+Shift+Space` (`Ctrl+Shift+Space` on Windows) or click **record** on the overlay pill.
While recording, the pill shows a live indicator. The tray tooltip switches to “Raven (Recording)”. On Windows 11 a new tray icon may start in the `^` overflow.
## What gets captured
| Stream | Source | Transcript label |
| ---------------- | ----------------------------------------------------------------- | ---------------- |
| **Microphone** | Your voice (AEC-cleaned, then residual-gated) | **You** |
| **System audio** | Speakers / meeting playback (WASAPI loopback or ScreenCaptureKit) | **Them** |
Live capture uses the **OS default** mic and playback devices. The Settings mic picker is only for the in-app mic **test**.
You can turn off system-audio capture in Settings (`captureSystemAudio`). Then only the mic stream runs.
## Echo cancellation
Without AEC, speaker audio leaks into the mic and shows up as **You**.
1. System PCM is the GStreamer **reference** (`webrtcechoprobe`).
2. Mic PCM goes through `webrtcdsp` (WebRTC AEC3).
3. `ResidualEchoGate` compares raw mic to recent system audio and drops leftover speaker bleed before STT.
If the AEC pipeline stalls or overflows, Raven can temporarily bypass AEC so capture continues.
## Live transcription
Two STT connections run in parallel (mic → You, system → Them).
* **Interim** text is italic while someone is still speaking
* **Final** text replaces interims on pause
* Labels stay **You** / **Them** (or your display name for You)
### Engine pick
Settings **`sttProvider`**: `auto` (default), `assemblyai`, or `deepgram`. You need at least one STT key or recording will not start.
| Preference | What happens |
| -------------- | ---------------------------------------------------------------------------------------------------------------------------------- |
| **auto** | AssemblyAI `u3-rt-pro` for English, Spanish, French, German, Portuguese, Italian when that key exists; otherwise Deepgram `nova-3` |
| **assemblyai** | AssemblyAI when the language is supported and keyed; otherwise Deepgram if keyed |
| **deepgram** | Deepgram `nova-3` when keyed |
`multi` / auto-detect language uses Deepgram. AssemblyAI failures can fall back to Deepgram.
### Languages
Set transcription language in **Settings → Language**. Deepgram `nova-3` covers a wide set including multilingual auto-detect. AssemblyAI Universal-3 Pro realtime covers the six languages above.
## Stopping
Press `Cmd/Ctrl+Shift+Space` again or click stop. Raven:
1. Stops capture
2. Saves transcript and overlay chat (unless incognito)
3. Generates a title and summary with a cheap/fast model on your LLM key
Nothing syncs to a Raven cloud.
## Permissions
### macOS
| Permission | Why |
| -------------------- | --------------------------------- |
| **Microphone** | Your voice |
| **Screen Recording** | System audio via ScreenCaptureKit |
Grant them in **System Settings → Privacy & Security** if the first prompt was denied.
### Windows
Uses WASAPI. If capture is silent, set the correct **default** playback and recording devices in Windows Sound settings.
# Sessions
Source: https://docs.useraven.ai/features/sessions
Local SQLite history, search, incognito, summaries, and insights.
A **session** is one recording: transcript, overlay chat, mode, and metadata. Non-incognito sessions live in SQLite on this machine (`data/raven.db` under the app user-data folder). Nothing syncs to a Raven cloud.
## History
After you stop, the dashboard shows:
* AI-generated **title**
* **Duration**
* Active **mode**
* **Timestamp**
## Detail tabs
| Tab | Content |
| -------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Summary** | Markdown summary (editable, regenerate with your key), plus [action items, a follow-up email draft, and talk ratio](/features/meeting-intelligence) |
| **Transcript** | You / Them with timestamps |
| **Usage** | Overlay questions and answers from the call |
| **Insights** | On-demand sentiment, topics, and key phrases with your LLM key |
| **Ask** | [Chat with this meeting](/features/ask) - streamed and saved per session |
Insights are not a hosted Pro feature. They call the same providers you configured in Settings.
The header also has an **Export** menu (Markdown or PDF, optional transcript) - see [Post-call recap](/features/meeting-intelligence).
## Copy
* Summary as markdown
* Transcript: yours, theirs, or full
* Individual usage cards
* Insight sections
## Search
Dashboard search matches titles, summaries, and transcript text.
## Auto-save
During a recording, Raven writes the session about every **60 seconds**. After a crash, an in-progress session can reopen as **Recovered Session**.
## Incognito
Toggle the hat-and-glasses control on the pill **before** you start. Incognito sessions are not written to SQLite and disappear when you stop.
## Reset local data
Packaged app folder is **Raven**. `npm run dev` uses **project-raven**.
```bash theme={null}
rm -rf ~/Library/Application\ Support/Raven/
rm -rf ~/Library/Application\ Support/project-raven/
```
```powershell theme={null}
rmdir /s /q "$env:APPDATA\Raven"
rmdir /s /q "$env:APPDATA\project-raven"
```
# Updating Raven
Source: https://docs.useraven.ai/features/updates
How Raven keeps itself current on macOS and Windows - in-app, keys and history intact.
Raven checks [GitHub Releases](https://github.com/Laxcorp-Research/project-raven/releases) hourly for a newer version and updates in place. Your API keys and local session history are untouched by an update.
## In-app update
When a newer version is published you'll see a banner (and a **Check for updates** button in **Settings → General**):
Raven downloads the new version in the background and shows progress.
Click to quit and install. Raven reopens on the new version.
## Platform notes
| Platform | How it updates |
| ---------------- | -------------------------------------------------------------------------------------------------------------- |
| **Windows** | `electron-updater` (NSIS). Downloads, then installs on restart. |
| **macOS 2.4.1+** | Builds are signed + notarized, so macOS self-updates the same way - **Update now**, then **Restart & update**. |
Updating **to** 2.4.1 on a Mac still uses the old flow one last time (download the installer, drag to Applications), because your installed 2.4.0 predates in-app self-update. From **2.4.1 onward**, macOS updates in place like Windows.
Raven only offers a build when the published version is newer than the one you're running.
# Introduction
Source: https://docs.useraven.ai/introduction
Open-source meeting copilot with dual-stream capture, local echo cancellation, and BYOK transcription and AI.
Raven is an Electron desktop app. It captures microphone and system audio, cancels echo on your machine, transcribes **You** and **Them** on two parallel streams, and answers from the live transcript. You bring your own API keys.
This product has **no Raven account, no hosted backend, and no cloud session sync**. Capture, echo cancellation, SQLite history, and document RAG run locally. Audio goes to **Deepgram** or **AssemblyAI**. Assist prompts go to **Anthropic** or **OpenAI**.
## Get Raven
Prebuilt installers for Windows 10/11 (x64) and macOS 12+ (Apple Silicon). Enter API keys on first launch.
Clone the repo and compile native capture + AEC yourself. Intel Macs need this path.
**macOS** installers are signed + notarized - just double-click to open (macOS shows the usual "downloaded from the Internet" prompt once). **Windows** OSS builds are unsigned, so SmartScreen may warn - choose **More info → Run anyway**. Raven [updates itself in-app](/features/updates) and only offers a build when the published version is newer than yours.
## Features
Mic + system audio. macOS: ScreenCaptureKit + CoreAudio. Windows: WASAPI loopback + capture (Rust/NAPI). Live recording uses the OS default devices.
GStreamer WebRTC AEC3 (`webrtcechoprobe` / `webrtcdsp`), then a residual echo gate so leftover speaker bleed does not land in **You**.
Two STT streams. Deepgram `nova-3` and/or AssemblyAI `u3-rt-pro`. Auto-routing prefers AssemblyAI for English, Spanish, French, German, Portuguese, and Italian when that key is present.
Anthropic or OpenAI from the overlay - Assist, recap, follow-up, custom prompts. Optional screenshot via `desktopCapturer`.
Electron `setContentProtection` so the overlay and dashboard are omitted from typical screen-share APIs. Not a guarantee against every capture tool.
Upload `.txt`, `.md`, `.pdf`, or `.docx`. Embedded on-device with MiniLM. Chunks stay in SQLite and are injected into Assist.
During a recording, Assist keeps recent turns and may compress older context in RAM. There is no cross-meeting user-memory profile.
Saved locally in SQLite with title, summary, transcript, and recap. Insights use your LLM key. Incognito skips persistence.
Chat with one meeting or across your whole history. On-device retrieval, your model, streamed answers with cited sources.
Action items, a one-click follow-up email, talk ratio, and Markdown / PDF export - all from your own key.
Detect a Zoom, Meet, Teams, or Webex meeting and offer to start recording. No bot ever joins.
## How it works
Overlay record button or `Cmd/Ctrl + Shift + Space`. Incognito, if on, will not write the session to SQLite.
System audio and microphone are captured together. macOS needs Microphone and Screen Recording. Windows uses the default devices.
System PCM is the AEC reference. Mic PCM goes through GStreamer, then `ResidualEchoGate` drops leftover speaker echo before **You** STT.
Two connections: mic → You, system → Them. Engine pick is Settings `sttProvider` plus language. AssemblyAI can fall back to Deepgram.
`Cmd/Ctrl + Enter` builds a prompt from the mode brief, RAG chunks, session memory, transcript tail, and an optional screenshot, then streams from your LLM.
## Platform support
| Platform | Status |
| --------------------------- | ----------------------------------------------------- |
| **Windows 10/11** (x64) | Supported - [Setup.exe](https://useraven.ai/download) |
| **macOS 12+** Apple Silicon | Supported - [DMG](https://useraven.ai/download) |
| **macOS Intel** | Build from source |
| Linux | Not supported (no native capture path) |
## Keyboard shortcuts
These match `registerGlobalHotkeys` in the app. On Windows use `Ctrl` instead of `Cmd`.
| Action | Shortcut |
| -------------------- | ------------------------- |
| Toggle overlay | `Cmd + \` |
| AI Assist | `Cmd + Enter` |
| Start/Stop recording | `Cmd + Shift + Space` |
| Clear conversation | `Cmd + Shift + Backspace` |
| Move overlay | `Cmd + Arrow Keys` |
| Scroll overlay | `Cmd + Shift + Up/Down` |
Recording used to be `Cmd+R` and clear used to be `Cmd+Shift+R`. Those stole browser refresh globally, so they were changed.
## Next steps
Capture, AEC, and STT routing.
Assist, modes, and local documents.
Chat with your meetings, action items, follow-up email, export.
Native dependencies and `npm run dev`.
How the processes and natives fit together.
# macOS Setup
Source: https://docs.useraven.ai/quickstart/macos
Complete macOS setup guide from a fresh machine to a running app.
Tested on macOS 12 (Monterey) through macOS 15 (Sequoia), Intel and Apple Silicon.
```bash theme={null}
xcode-select --install
```
A system dialog will appear - click **Install** and wait for it to finish (\~2 min).
Verify:
```bash theme={null}
xcode-select -p
# Expected: /Library/Developer/CommandLineTools (or an Xcode.app path)
```
If you see `xcode-select: error: command line tools are already installed` - you're good, move on.
Install via [nvm](https://github.com/nvm-sh/nvm) (recommended). Skip the `curl` line if you already have nvm.
```bash theme={null}
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.1/install.sh | bash
```
**Close and reopen your terminal**, then:
```bash theme={null}
nvm install 22
nvm use 22
```
Verify:
```bash theme={null}
node -v
# Expected: v22.x.x (any 22+ version)
```
If `nvm: command not found`: Close your terminal and open a new one - nvm's install script adds itself to your shell profile, but only new shells pick it up.
```bash theme={null}
brew install gstreamer gst-plugins-base gst-plugins-good gst-plugins-bad
```
Don't have Homebrew? Install it first from [brew.sh](https://brew.sh).
Verify:
```bash theme={null}
pkg-config --modversion gstreamer-1.0
# Expected: 1.24.x (or similar)
```
Homebrew's `pkg-config` path isn't set. Add the correct line to your `~/.zshrc` and restart your terminal:
```bash theme={null}
# Apple Silicon (M1/M2/M3/M4):
echo 'export PKG_CONFIG_PATH="/opt/homebrew/lib/pkgconfig:$PKG_CONFIG_PATH"' >> ~/.zshrc
# Intel Mac:
echo 'export PKG_CONFIG_PATH="/usr/local/lib/pkgconfig:$PKG_CONFIG_PATH"' >> ~/.zshrc
```
```bash theme={null}
git clone https://github.com/Laxcorp-Research/project-raven.git
cd project-raven
npm install
```
`npm install` takes a few minutes. It automatically rebuilds `better-sqlite3` for Electron via the `postinstall` script - you'll see `@electron/rebuild` output near the end.
Verify:
```bash theme={null}
ls node_modules/.package-lock.json && echo "OK"
# Expected: OK
```
If `npm install` fails with `node-gyp` errors, make sure Xcode Command Line Tools installed successfully in Step 1. Run `xcode-select -p` to confirm.
```bash theme={null}
cd src/native/aec
npm install
./build-deps.sh
npx cmake-js compile
cd ../../..
```
What this does:
1. Installs the addon's build tools (`cmake-js`, `node-addon-api`)
2. Verifies all GStreamer libraries and builds the WebRTC DSP plugin from source (Homebrew doesn't ship it)
3. Compiles the C++ echo-cancellation native module
Verify:
```bash theme={null}
ls src/native/aec/build/Release/raven-aec.node && echo "OK"
# Expected: OK
```
* **`build-deps.sh` fails with "gstreamer-1.0 not found"**: Revisit Step 3 and make sure `pkg-config --modversion gstreamer-1.0` works.
* **`cmake-js compile` fails with "cmake not found"**: cmake is bundled with cmake-js. Run `npx cmake-js --version` - if that fails, delete `node_modules` inside `src/native/aec/` and re-run `npm install`.
```bash theme={null}
cd src/native/swift/AudioCapture
swift build -c release
cd ../../../..
```
Verify:
```bash theme={null}
ls src/native/swift/AudioCapture/.build/release/audiocapture && echo "OK"
# Expected: OK
```
Your Swift toolchain may be too old (5.9+ required). Check with `swift --version`. Update Xcode Command Line Tools:
```bash theme={null}
sudo rm -rf /Library/Developer/CommandLineTools && xcode-select --install
```
```bash theme={null}
npm run dev
```
The Electron app opens. On first launch enter a transcription key (Deepgram and/or AssemblyAI) and an AI key (Anthropic or OpenAI).
If audio capture doesn't work: macOS requires explicit permissions. Go to **System Settings > Privacy & Security** and grant both **Microphone** and **Screen Recording** access to the app (or to your terminal emulator during development).
## Required Permissions
macOS requires explicit user consent for audio capture. On first recording, the app will request these:
| Permission | Why it's needed | How to grant |
| -------------------- | -------------------- | ------------------------------------------------------- |
| **Microphone** | Capture your voice | System Settings > Privacy & Security > Microphone |
| **Screen Recording** | Capture system audio | System Settings > Privacy & Security > Screen Recording |
During development, you may need to grant these permissions to your **terminal emulator** (e.g., iTerm, Terminal.app) rather than the Electron app directly. In production builds, the signed `.app` is the target.
## Troubleshooting
| Symptom | Fix |
| ----------------------------------------------- | ------------------------------------------------------------------------------------------------------------------- |
| `build-deps.sh`: "gstreamer-1.0 not found" | Revisit Step 3 - check `pkg-config --modversion gstreamer-1.0` |
| `cmake-js compile` fails with "cmake not found" | Run `npx cmake-js --version` - if it fails, delete `node_modules` inside `src/native/aec/` and re-run `npm install` |
| `swift build` fails with unresolved imports | Swift toolchain too old (5.9+ required) |
| `npm install` fails with `node-gyp` errors | `xcode-select --install` - ensure CLI tools are installed |
| `NODE_MODULE_VERSION mismatch` at runtime | `npx @electron/rebuild -f -w better-sqlite3` from the project root |
| No audio captured despite permissions granted | Restart the app - macOS sometimes requires a restart after granting Screen Recording |
| App starts, no audio | System Settings > Privacy & Security: grant Microphone and Screen Recording |
# Build from Source
Source: https://docs.useraven.ai/quickstart/overview
Set up the open-source version of Raven on your machine.
Prefer a prebuilt installer? Use the [download page](https://useraven.ai/download). This page is only for compiling from source.
Build Raven locally and enter your own API keys in the app. Pick your platform and follow each step in order.
## Prerequisites
* [Deepgram](https://console.deepgram.com) and/or [AssemblyAI](https://www.assemblyai.com) - real-time transcription
* [Anthropic](https://console.anthropic.com) or [OpenAI](https://platform.openai.com) - AI assistance
* After onboarding, open **Settings → API Keys** to switch STT engine, provider, and model
* **Node.js** 22.12.0 or higher
* **GStreamer** for echo cancellation
* Platform-specific build tools (Xcode CLI on macOS, Visual Studio Build Tools on Windows)
## Choose Your Platform
macOS 12 (Monterey) through macOS 15 (Sequoia), Intel and Apple Silicon.
Windows 10 (21H2+) and Windows 11. All commands for PowerShell.
## Testing
After setup, verify everything works:
```bash theme={null}
npm test # Unit + integration tests
npm run test:coverage # With coverage report
npm run test:e2e # End-to-end (requires npm run build first)
npm run test:all # Everything
```
## Troubleshooting
| Symptom | Likely Cause | Fix |
| ----------------------------------------- | ---------------------------------------------- | --------------------------------------------------------------------------- |
| `NODE_MODULE_VERSION mismatch` at runtime | Native module built for wrong Electron version | `npx @electron/rebuild -f -w better-sqlite3` |
| App starts, no audio on macOS | Missing system permissions | System Settings > Privacy & Security: grant Microphone and Screen Recording |
| App starts, no audio on Windows | Wrong default audio device | Settings > Sound: set correct default playback/recording devices |
**Reset all data (fresh start):**
```bash theme={null}
rm -rf ~/Library/Application\ Support/Raven/
rm -rf ~/Library/Application\ Support/project-raven/
```
```powershell theme={null}
rmdir /s /q "$env:APPDATA\Raven"
rmdir /s /q "$env:APPDATA\project-raven"
```
Packaged builds use **Raven**. `npm run dev` uses **project-raven**.
# Windows Setup
Source: https://docs.useraven.ai/quickstart/windows
Complete Windows setup guide from a fresh machine to a running app.
Tested on Windows 10 (21H2+) and Windows 11. All commands are for **PowerShell**. Open a **new terminal** after each installer to pick up PATH changes.
Download and run the [Visual Studio Build Tools installer](https://visualstudio.microsoft.com/visual-cpp-build-tools/).
In the installer, check the **"Desktop development with C++"** workload and click Install. Make sure these optional components are selected (they should be by default):
* MSVC Build Tools for x64/x86 (Latest)
* Windows 10/11 SDK
* C++ CMake tools for Windows
Verify:
```powershell theme={null}
& "${env:ProgramFiles(x86)}\Microsoft Visual Studio\Installer\vswhere.exe" -products * -requires Microsoft.VisualStudio.Workload.VCTools -property displayName
# Expected: Visual Studio Build Tools 2022
```
If you have full Visual Studio (not just Build Tools) with the C++ workload, that works too.
**Option A** - [nvm-windows](https://github.com/coreybutler/nvm-windows/releases) (recommended):
Download and run the latest `nvm-setup.exe`, then open a **new** terminal:
```powershell theme={null}
nvm install 22
nvm use 22
```
**Option B** - Download the LTS 22.x MSI installer from [nodejs.org](https://nodejs.org/).
Verify (in a **new** terminal):
```powershell theme={null}
node -v
# Expected: v22.x.x
```
The project requires `node >= 22.12.0` (see `package.json` engines). Using `nvm install lts` may install a newer major version that hasn't been tested.
Python is required by `node-gyp` to compile native Node.js modules.
```powershell theme={null}
winget install Python.Python.3.12 --source winget
```
Or download from [python.org](https://www.python.org/downloads/) - make sure "Add to PATH" is checked.
Verify (in a **new** terminal):
```powershell theme={null}
python --version
# Expected: Python 3.x.x
```
Download and run [rustup-init.exe](https://rustup.rs/). Accept the defaults (installs `stable-msvc`).
Verify (in a **new** terminal):
```powershell theme={null}
rustc --version
# Expected: rustc 1.xx.x (...)
rustup default stable-msvc
```
Download the **MSVC x86\_64** installer from [gstreamer.freedesktop.org/download](https://gstreamer.freedesktop.org/download/) - click **Windows > MSVC x86\_64 (VS 2022, Release CRT)**.
For GStreamer 1.28+, there is a single combined installer (runtime + development). For older versions, download both the Runtime and Development MSI files.
Run with default settings. After installation, verify (open a **new** terminal):
```powershell theme={null}
echo $env:GSTREAMER_1_0_ROOT_MSVC_X86_64
# Expected: C:\gstreamer\1.0\msvc_x86_64\ (or similar)
```
The installer didn't set it. Find where GStreamer was installed and set it manually:
```powershell theme={null}
[Environment]::SetEnvironmentVariable("GSTREAMER_1_0_ROOT_MSVC_X86_64", "C:\Program Files\gstreamer\1.0\msvc_x86_64\", "User")
```
Then **restart your terminal**.
If GStreamer installed to `C:\Program Files\gstreamer\` instead of `C:\gstreamer\`, that's fine - just make sure the path matches your installation.
```powershell theme={null}
winget install Kitware.CMake --source winget
```
Or download from [cmake.org/download](https://cmake.org/download/) - make sure "Add to PATH" is checked.
Verify (in a **new** terminal):
```powershell theme={null}
cmake --version
# Expected: cmake version 3.x.x
```
```powershell theme={null}
git clone https://github.com/Laxcorp-Research/project-raven.git
cd project-raven
npm install
```
`npm install` takes a few minutes. It automatically rebuilds `better-sqlite3` for Electron via the `postinstall` script.
Verify:
```powershell theme={null}
Test-Path node_modules\.package-lock.json
# Expected: True
```
```powershell theme={null}
# Fix 1: Set the version hint for node-gyp
npm config set msvs_version 2022
Remove-Item -Recurse -Force node_modules
npm install
# Fix 2: Environment variable (works on all npm versions)
$env:GYP_MSVS_VERSION = "2022"
Remove-Item -Recurse -Force node_modules
npm install
```
First, check the Electron version:
```powershell theme={null}
node -e "console.log(require('./node_modules/electron/package.json').version)"
# Note the version (e.g. 40.4.1)
```
Then build targeting that version:
```powershell theme={null}
cd src\native\aec
npm install
npx cmake-js compile --runtime electron --runtime-version
cd ..\..\..
```
Replace `` with the version from the previous command.
The `--runtime electron --runtime-version` flags are **required**. Without them, the addon is built for Node.js instead of Electron and will **crash** when loaded. If you upgrade Electron later, you must rebuild this addon with the new version.
The `build-deps.sh` script is macOS-only. On Windows, the GStreamer MSVC installer already includes all required plugins (including WebRTC DSP).
Verify:
```powershell theme={null}
Test-Path src\native\aec\build\Release\raven-aec.node
# Expected: True
```
```powershell theme={null}
cd src\native\windows
npm install
npx napi build --platform --release
cd ..\..\..
```
Verify:
```powershell theme={null}
Test-Path src\native\windows\raven-windows-audio.win32-x64-msvc.node
# Expected: True
```
* **Linker errors**: Make sure Rust is using the MSVC target: `rustup default stable-msvc`
* **"Windows SDK not found"**: Open Visual Studio Installer > Modify > Individual components and install the latest "Windows 10 SDK" or "Windows 11 SDK"
```powershell theme={null}
npm run dev
```
The Electron app opens. On first launch you'll see a 6-step onboarding flow - enter a transcription key (Deepgram and/or AssemblyAI) and an AI key (Anthropic or OpenAI). WASAPI uses the OS default devices; the Settings mic picker is only for the in-app mic test.
If audio capture doesn't work, check **Settings > Sound** and make sure the correct playback and recording devices are set as default. WASAPI captures from the default devices.
## Troubleshooting
| Symptom | Fix |
| ----------------------------------------------- | --------------------------------------------------------------------------------- |
| `Could not find any Python installation` | Install Python 3.x and add to PATH (Step 3) |
| `Could not find any Visual Studio installation` | Set `$env:GYP_MSVS_VERSION = "2022"`, delete `node_modules`, re-run `npm install` |
| `NODE_MODULE_VERSION mismatch` at runtime | `npx @electron/rebuild -f -w better-sqlite3` from the project root |
| cmake-js: "CMake is not installed" | Install CMake (Step 6) |
| cmake-js: "GStreamer not found" | Set `GSTREAMER_1_0_ROOT_MSVC_X86_64` and restart terminal (Step 5) |
| AEC addon crashes Electron on startup | Rebuild with `--runtime electron --runtime-version` (Step 8) |
| `napi build` linker errors | `rustup default stable-msvc` and ensure VS Build Tools C++ workload is installed |
| App starts, no audio | Settings > Sound: set correct default playback/recording devices |