Reference Docs
MCP access is how an AI assistant outside FocusAnalyze — Claude, an assistant in your editor, an MCP-capable tool — reads and acts on your real FocusAnalyze data.
For per-client setup instructions, see MCP servers. This page covers the access itself: what a token is, and what it's allowed to do.
Settings → MCP Access. Generating gives you two things: the MCP server URL, and a token.
The token is shown once. Copy it when it's displayed; if you lose it, revoke it and generate a new one rather than trying to recover it.
A fresh token is read-only. A connected assistant can see your tasks, projects, events, habits, reports and journals, and can answer questions about them — but it cannot change anything.
This is the default on purpose. Most of what you'd want an assistant to do is read.
Write access is granted per project, not globally, in the same settings section. An assistant with a write grant on one project can create and update tasks there and nowhere else.
If an assistant reports a permission error while trying to create or update something, this is almost always why: the token is connected and working, but that project has no write grant.
Deleting is not possible through MCP at all. There are no delete tools, for any token, with any grant.
Revoke a token in the same place. Revocation takes effect immediately and disconnects any assistant using it.
Your personal preferences and FocusAI's long-term memory are deliberately outside MCP access — no token can read them.
The VS Code and Cursor extension signs in on its own and doesn't use these tokens. The two are independent: connecting one doesn't connect the other.