Connect Cursor to your fenbs board

Connect

1. In Cursor, open the MCP settings (or edit mcp.json directly — in the project under .cursor/, or in your home folder for every project).
2. Add fenbs as a server with the URL below.
3. Cursor prompts you to authenticate: a browser opens on fenbs, you sign in, tick what it may do, and approve.
4. In the agent chat, ask what is on your board. The first call is fenbs_whoami.

{
  "mcpServers": {
    "fenbs": {
      "url": "https://fenbs.ai/api/mcp"
    }
  }
}

A project-level file keeps a board next to the code it belongs to: the agent working in that repository talks to that board. A machine that cannot open a browser can send a token from Settings, “Connect an AI assistant”, as an Authorization: Bearer header instead.

What it can do once connected

• Read the board: fenbs_list_items with filters by lane, kind or free text, and fenbs_get_item for one task with its comments.
• Add tasks with fenbs_create_item — a Feature, an Enhancement or a Bug, in a lane, with a priority.
• Move and edit with fenbs_update_item: change the lane, the title, the note, the kind or the priority.
• Comment with fenbs_comment, and search across everything it can see with fenbs_search.
• Read your AI context with fenbs_get_context — the notes you keep on the board about how you work — and add to it with fenbs_add_context_note.
• Ask fenbs_whoami first: it answers who the assistant is acting as and what it may do, which is the question to settle before it writes anything.

What it cannot do

Anything you cannot. The assistant acts as you and holds your role on the board, narrowed by the scopes you ticked when you approved it: read, write, comment. If your role cannot move tasks between lanes, neither can the assistant, and it is told so in a sentence it can repeat to you: “You are a Viewer on Kitchen Refit, which cannot move tasks between lanes.” Revoke its token in Settings and it stops at once; everything it did stays in the history under its name.

A working rhythm

Put a line in your Cursor rules: before starting, read fenbs AI context and the In Progress lane; after finishing, comment on the task and move it. The board becomes the record of what the agent did, in words you can read, without a transcript.