← Back

How to use Canvas MCP

Connect your Canvas LMS account to AI assistants using connectors. Follow the steps below for Claude or ChatGPT.

Using Claude (Connectors)

Claude supports Connectors so you can add external tools and data. Add Canvas MCP as a Custom Connector; Claude will use OAuth to open a browser where you enter your Canvas domain and token.

  1. In Claude, open Settings (gear icon) and go to Connectors.
  2. Click Add Custom Connector and enter this connector URL:
    https://canvas-ghcawuze4-inflectlabs.vercel.app/mcp
  3. Claude will discover the OAuth metadata and, when you use the connector, open a browser window to this site’s setup page.
  4. On the setup page, enter your Canvas domain (e.g. https://canvas.sydney.edu.au) and your Personal Access Token (from Canvas → Account → Settings → New Access Token).
  5. Click Test Connection, then Connect. After the redirect back to Claude, you can ask Claude about your courses, assignments, and grades.

Your credentials are stored encrypted and only used to call the Canvas API on your behalf; they are never sent to Claude.

Using ChatGPT (Custom GPT and Actions)

ChatGPT connects to external APIs via GPT Actions. You create a Custom GPT and add an Action that describes your API using an OpenAPI schema. ChatGPT then uses that schema to call your API when answering the user.

  1. Create a Custom GPT (ChatGPT → Explore GPTs → Create).
  2. In the GPT editor, go to the Actions section and add a new action. You’ll need an OpenAPI schema that describes the Canvas API (or a proxy) your GPT will call.
  3. Configure Authentication: for APIs that need a token, use either API Key or OAuth in the Action settings. If the server supports OAuth, add the callback URL that ChatGPT shows you into your app.
  4. In the Custom GPT Instructions, describe when and how to use the action (e.g. “When the user asks about courses or assignments, call the relevant API action and use the result to answer”).
  5. Test the action with the built-in test button, then try representative questions to confirm the GPT calls the right endpoints with the right parameters.

Canvas MCP currently exposes tools via the Model Context Protocol (MCP) for Claude and other MCP clients. To use Canvas with ChatGPT in the same way, the server would need to expose an OpenAPI-compatible API and schema; that integration may be added in a future release.