Course Bindery
Connect your agent
Course Bindery speaks the Model Context Protocol, so Claude — or any other MCP client — can draft lessons, edit blocks and publish courses in your workspace. We host no model and bill nothing for it: the agent is yours, on your own subscription.
The server URL
https://api.coursebindery.com/mcp/Paste it with no header. The client discovers the rest — it registers itself and sends you here to approve what it may do.
Per client
- claude.ai
- Customize → Connectors → Add, and paste the URL. Claude works the rest out for itself.
- Claude Desktop
- Settings → Connectors → Add a remote server, and paste the same URL.
- Claude Code
- claude mcp add --transport http learn https://api.coursebindery.com/mcp/
- Cursor
- .cursor/mcp.json → {"mcpServers":{"learn":{"url":"https://api.coursebindery.com/mcp/"}}}
Claude Code and Cursor can also use a workspace API key from Settings → Agents, which suits automation and CI. Prefer signing in where the client supports it: the token is then issued to you, so what the agent writes is attributed to you and you can disconnect it on its own.
What an agent may do
An agent is granted an author’s reach and no more — never your workspace settings, your webhooks, or the ability to sign learners in. A grant also never exceeds your own role: if you are a viewer, an agent connected by you can only read.
- Read your courses
courses:read - Create and edit courses
courses:write - Publish courses to learners
courses:publish - Upload images and files
media:write - Read your workspace settings
org:read - See your learners and their progress
learners:read - Read learner progress
progress:read
Every call is audited with who made it and through which client, and every write is a revision you can inspect or restore. Settings → Agents lists what is connected and revokes any of it.