MCP layer¶

On top of the API there is an MCP layer for your AI workflows. MCP, the Model Context Protocol, is the interface through which AI tools talk to other applications.

What AI tools can do with it¶

• Read: topics, posts, decisions and the ODR register.
• Write: create topics and posts themselves and prepare ODR drafts.

The tool acts as the person who connected it. It sees exactly what that person sees and may do exactly what that person may do.

First: enable access¶

Access is switched off per workspace until someone switches it on. Workspace managers enable both API access and MCP access under Settings → Integrations. When signing in, Rhizome offers only workspaces where both are enabled.

Connecting¶

The MCP server is at:

https://api.get-rhizome.app/mcp

Every AI tool signs in through Rhizome in the browser: you sign in, choose the workspace and whether the tool may only read or also write, and allow access. You don’t need a key for this.

This is how you add the server in common tools:

If you use Claude Code, the easiest way to connect the server is the plugin, which also brings the skills.

Afterwards you can ask your AI tool whether Rhizome works. The setup skill checks the connection and tells you what to fix.

Connected apps and switching workspace¶

You see your connected tools in Rhizome under Security and Access → Connected apps. There you can disconnect any of them. To switch workspace, sign in again from the tool and choose the other one.

Labeled “via MCP” or “via API”¶

What an AI tool creates appears in Rhizome under the name of the person who connected it. So that nobody has to guess who wrote it, such topics and posts are labeled “via MCP” or “via API”.

Writing with confirmation¶

Everything an AI tool writes lands in your name and may notify others. The Rhizome skills therefore show a draft first and write only after your yes. They never publish ODRs themselves: they create a draft that you read and publish in the app.