Canvas Chat

API Keys

Keys are stored locally in your browser and sent directly to providers.

OpenAI API Key

Anthropic API Key

Google AI API Key

Groq API Key

GitHub Token (for GitHub Models) PAT with models:read scope. Create token

GitHub Copilot Models Copilot models (e.g., github_copilot/gpt-4) require device authentication. On first use, check your server terminal for a device code and URL to authenticate with GitHub.

Exa API Key (for web search) Use /search query to search the web. Get API key

LLM Proxy

Configure a custom base URL to use your own LLM proxy (e.g., LiteLLM proxy, OpenRouter).

Base URL (optional) Leave empty to use default provider endpoints. Supports OpenAI-compatible APIs.

Custom Models

Add LiteLLM-compatible models for self-hosted or proxy endpoints.

Model ID * Format: provider/model-name (e.g., openai/my-model, ollama_chat/llama3.1)

Display Name Optional. Defaults to model ID if not specified.

Context Window

Base URL

Flashcard Settings

Configure how flashcard answers are graded during review.

Grading Strictness Controls how strictly the AI grades your flashcard answers

Getting Started

Type a message and press Enter to start a conversation
Click on a node to select it, then type to reply
Ctrl/Cmd + click to select multiple nodes
Drag the canvas to pan, scroll to zoom
Double-click empty space to fit all nodes in view

Keyboard Shortcuts

`Ctrl`/`Cmd` + `K`	Search nodes
`Ctrl`/`Cmd` + `Z`	Undo
`Ctrl`/`Cmd` + `Shift` + `Z`	Redo
`Delete` / `Backspace`	Delete selected nodes
`R`	Focus reply input (when node selected)
`C`	Copy node content (when node selected)
`F`	Fit node to viewport (when node selected)
`Escape`	Clear selection / Close dialogs
`Enter`	Send message
`Shift` + `Enter`	New line in message
`?`	Show this help

Slash Commands

`/note` text or URL	Create a note (supports PDF URLs)
`/search` query	Search the web with Exa AI
`/research` topic	Deep research with multiple sources
`/matrix` context	Create a comparison matrix
`/committee` question	Consult multiple LLMs and synthesize

Tips

Branching: Select text in a node to create a branch from that specific point
Semantic zoom: Zoom out to see node summaries instead of full content
Context: The context budget bar shows how much of the model's context window is used
Tags: Use the 🏷️ button to organize nodes with colored tags
Stop generation: Click "⏹ Stop" on an AI response to stop and optionally continue later
PDF import: Click 📎 to upload a PDF, or drag & drop onto the canvas

Canvas Chat

Tags

Settings

API Keys

LLM Proxy

Custom Models

Flashcard Settings

Sessions

Create Matrix

Rows

Columns

Edit Matrix

Rows

Columns

Cell Details

Row Details

Help

Getting Started

Keyboard Shortcuts

Slash Commands

Tips

Edit Content

LLM Committee

Edit Title

Generate Flashcards

Review Flashcards 1/5

Select Claims to Verify