Why Canvas MCP
Canvas LMS exposes a rich REST API, but wiring AI agents to it requires token provisioning, session management, and CSRF handling — plumbing that most teams would rather skip. Canvas MCP handles all of that automatically. Your agent connects to the MCP server; the server reads live cookies straight from Chrome on every request, so session and CSRF rotation are picked up without any extra work.Key features
Zero-token auth
Reads Canvas session cookies directly from Chrome. No API token generation, no
.env secrets to rotate.FastMCP server
Exposes 25+ Canvas tools over stdio, SSE, or HTTP transport so any MCP-compatible client can connect.
Typed CLI
The
canvas command lets you list courses, inspect assignments, read discussions, and check auth status from the terminal.Auto domain detection
If
CANVAS_BASE_URL is unset, the CLI infers your Canvas domain from Chrome cookies automatically.Multiple transports
Supports stdio (desktop clients), HTTP (networked agents), and SSE (legacy clients) out of the box.
Chrome profile support
Select which Chrome profile to read cookies from when you have multiple accounts.
How Chrome-based auth works
Canvas MCP uses thebrowser-cookie3 library to read your Canvas session cookies directly from Chrome’s local profile storage. It does not store credentials, does not proxy through a third party, and does not require you to export anything. On every request the server fetches fresh cookies, so if Canvas rotates your session the server stays in sync automatically.
On macOS, Chrome protects its cookie database with the system Keychain. The first time you run a Canvas MCP command, macOS will prompt you to allow access. Click Allow or Always Allow to proceed.
Next steps
Quickstart
Install the package and make your first Canvas query in under five minutes.
Installation
Full installation options, including the
uv tool install path and both CLI entrypoints.Auth overview
Learn how Chrome cookie auth works and how to configure your Canvas base URL.
MCP server overview
Understand transports, client configuration, and how to connect your AI agent.
