What an agent can do
An agent connected to TSO over MCP uses the same functions as the app:
- Report: query numbers as in the dashboard, in any attribution model and window, split by up to three dimensions. Look up single leads, without contact details unless you allow them.
- Set up: create projects, set the privacy mode, get the tracking and event code, count thank-you pages as leads and create custom events.
- Send back: link ad accounts, create Conversion API destinations, check them with a test lead and switch them live.
- Check: see in the diagnosis whether visits, leads and deliveries arrive, and send stuck deliveries again once the cause is fixed.
The tools walk the agent through the setup step by step: test mode first, then a test lead, then live. Changes take effect at once, as in the app.
A key with exactly the permissions it needs
You create keys in the workspace under "API & tracking domains". A template sets the scopes, and you can adjust them one by one:
- Analysis reads numbers and leads, without contact details.
- Setup agent can also create and set up projects, including ad accounts, destinations and their tokens.
- Everything adds the leads' contact details and reporting leads and events.
A key reaches all projects or only chosen ones, and can expire after 30 or 90 days. It is shown once and stored only as a hash. Revoke it, and every further request fails at once.
What stays with you
Sign-ins to ad platforms and integrations happen in your browser: the agent sends you the link, and the project tells it when the sign-in is there. An agent with its own scope can set tokens but never read them. Nothing is deleted over MCP or the API.
Connect
The MCP server runs at /mcp over Streamable HTTP, with the same keys and rate limit as the public API. Add the address and your key to your agent's settings. The guide is in the API documentation.