MCP and API
Everything you can do in Studio, an AI assistant or your own code can do too, within your access.
What you do in Studio is also available over MCP, the protocol AI assistants use to call tools. Every build checks that the two match.
An assistant acts as the person who connected it. It can only do what that person may do, and risky changes wait for their approval.
The same operations are available as an HTTP API. The API reference lists every operation, grouped by area, and marks the few that assistants cannot call, such as sign-in steps that need a browser.
Connect an AI assistant
In pilotConnect an AI assistant to Studio over MCP and sign in as yourself.
What you can do
- Connect an assistant with a browser sign-in
- See connected assistants and disconnect any of them
Approve what an assistant asks to do
In pilotWhen an assistant asks to do something risky, such as publish or send, it waits until you approve it in Studio.
What you can do
- Review a pending request with its details
- Approve or cancel it
Company MCP policy
In pilotAdministrators decide whether people may connect assistants, and what those assistants may do.
What you can do
- Allow or stop assistant connections
- Limit what connected assistants may do
HTTP API
In pilotThe same operations as MCP, over HTTP, each described with when to use it and who may call it.
What you can do
- Read the API reference
- Call operations as a signed-in person
Studio inside your own product
Coming nextShow parts of Studio inside your own app, so your users can shape their agents there.
What you can do
- Embed an agent's settings in your app