Skip to main content
The /schema endpoint returns every action available on a tracked process, along with each action’s type signature, parameter list, reversibility class, and semantic confidence score. Unlike GET /state, this endpoint does not return the objects map — it is a lightweight call designed for agent planning phases where you need to know what actions exist and how to call them, without the overhead of a full state capture.

Endpoint

Rate limit: 100 requests per second.

Parameters

pid
integer
required
The OS process ID of the target application. Obtain this from GET /apps.

Response Fields

pid
integer
required
The OS process ID that was queried.
app_id
string
required
Stable semantic identifier for the application.
app_name
string
required
Human-readable application name.
schema_version
string
required
Current semantic schema version in semver format, e.g. "1.4.2". Increments on every accepted POST /learn correction.
actions
object
required
A map of action names to ActionSchema objects. Each key is an action name you can pass to POST /action.

Examples

Example Response

Notes

  • The actions map returned by /schema is identical to the actions field in GET /state. Use /schema when you only need the action list and want to avoid the overhead of transferring the full objects map.
  • The Python SDK caches the schema on the session object. Calling session.actions does not make a network request after the first attach() or refresh().
  • If a new action appears (e.g., after navigating to a new view), call GET /state with fresh=true or GET /schema again — the session cache is invalidated automatically when schema_version changes.