Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.useraven.ai/llms.txt

Use this file to discover all available pages before exploring further.

Raven gives you an AI assistant that understands what’s happening in your meeting. It sees the full conversation transcript and can answer questions, suggest responses, summarize progress, and more.

​
AI Providers

Raven can use Anthropic (Claude) or OpenAI (GPT), depending on the edition:
EditionHow AI is configured
Raven ProKeys are managed for you. AI runs through Raven’s backend proxy (Claude). No API key or model dropdown in Settings.
Open sourceYou provide your own Anthropic or OpenAI API key and pick a model in Settings → API Keys.

​
Models (open source)

When you bring your own keys, you can choose:
ProviderOptions (fast → capable)
AnthropicClaude Haiku 4.5, Claude Sonnet 4.6, Claude Opus 4.7
OpenAIGPT-5.4 Mini, GPT-5.4, GPT-5.5, GPT-5.2
Haiku / GPT-5.4 Mini are recommended for live meetings; Opus 4.7 or GPT-5.5 are better for the hardest reasoning tasks (e.g. complex coding).

​
Quick Actions

During a recording, the overlay shows quick action buttons. Click any of them and the AI responds with full conversation context:
ActionWhat it does
AssistGeneral help based on the current conversation
What should I say?Suggests a response to the last thing the other person said
Follow-upProposes a relevant follow-up question
RecapSummarizes the conversation so far
On any AI response card, you can click Tell me more to expand on that answer. You can also trigger Assist with Cmd+Enter (includes a screenshot of your screen for visual context).

​
Typed Questions

Type a specific question in the input bar at the bottom of the overlay and press Enter. The AI answers your question directly, using the conversation transcript as context. This is useful for pointed questions like “What was their budget?” or “Summarize the technical requirements they mentioned.”

​
Screenshot Context

Press Cmd+Enter to capture a screenshot of your screen and send it to the AI along with your question. During a recording: The AI sees both the screenshot AND the conversation transcript. Useful for referencing slides, documents, or shared screens during a meeting. Without a recording: The AI operates purely on the visual content. This turns Raven into a general screen assistant - solve math problems, interpret charts, answer questions about anything visible on screen. The screenshot automatically excludes Raven’s own overlay, so the AI only sees what’s beneath it.

​
Math & Code Rendering

AI responses render with full formatting:
  • Markdown - bold, italic, bullet points, headers
  • Math formulas - rendered with KaTeX (e.g., equations from a whiteboard or slide)
  • Code blocks - syntax-highlighted with language detection

​
Fast vs Deep (Pro)

Raven Pro users see a Fast/Deep toggle on the overlay pill. There is no separate model picker in Settings — this toggle is how you control speed vs quality:
ToggleModelBest for
FastClaude Haiku 4.5Quick assists during live conversation — “what should I say?”, recap, follow-ups
DeepClaude Sonnet 4.6Harder questions that need more reasoning while still staying responsive
Pro routes all AI through Claude via the managed backend. For maximum reasoning (e.g. very hard coding problems), use the open-source build with your own key and select Claude Opus 4.7 in Settings, or stay on Deep in Pro for a strong default.

​
Copying Responses

Hover over any AI response card and click the copy icon to copy the response to your clipboard. You can paste it into notes, chat, or any other application.