FAQ

FAQ

Is this hosted? Does my data leave my machine?

No. takt-mcp runs locally and connects only to your Takt instance at TAKT_BASE_URL, using your API key. The data path is agent → takt-mcp → your Takt; nothing is hosted by us.

Can it modify my analytics?

No. It performs only GET requests against the Takt public read API. There are no write tools.

Why does list_sites return just one site?

Takt API keys are single-site bound. To cover several sites, mint one key per site and run one server instance per key (see Configuration → Multiple sites).

Which permissions does my key need?

stats:read for every reporting tool; sites:read for list_sites. The Tools page lists the permission for each tool.

Is my API key ever logged?

No. Even with TAKT_DEBUG=1, the key is never written to logs. Logs go to stderr only; stdout is the MCP transport.

What's the takt://sites resource?

When TAKT_ORG is set, the server exposes your org's sites as an MCP resource, so the agent can read site context without spending a tool call.

Which agents work with it?

Any MCP client: Claude Desktop, Claude Code, Cursor, and others. It speaks MCP over stdio.

How do custom date ranges work?

Pass from/to as YYYY-MM-DD (instead of period), plus an optional tz (IANA timezone) for bucketing. See the shared time filter in Tools.

Where do I report bugs or security issues?

Open an issue on the repo for bugs. For security, use the private advisory flow (see SECURITY.md) — never a public issue.