Skip to main content
See Filter Compatibility for the authoritative filter contract and the coverage of every MCP tool.
These tools read run status and credits. They change nothing and use no credits. Starting a run is a write operation: see Run write tool. They follow MCP read access and are available from Starter, without a write scope.

estimate_run

Estimate the credits a run would use, without reserving or debiting anything.

Parameters

Returns

An estimate is a snapshot. Starting the run revalidates the targets and the balance, so its result can differ.

OAuth scope

Requires: read:prompts

list_runs

List recent runs of a workspace with their status, from newest to oldest.

Parameters

Returns

Each run includes its purpose, the number of prompts and models, the requested and planned counts, a computed status, outcome counts and progressPercentage. The response also returns nextCursor (null when there is no further page), totalCount and activeCount for the whole workspace. Status is one of QUEUED, PROCESSING, COMPLETED, COMPLETED_WITH_WARNINGS, FAILED, CANCELLED or NO_WORK. Progress does not guarantee that every requested answer succeeded: read the status and the outcome counts.

OAuth scope

Requires: read:prompts

get_run

Get the status and counts of one run.

Parameters

Returns

Planning metadata for the run: its purpose, the prompt and model counts, the requested and planned counts, the number of recorded attempts, whether planning completed or failed, and timestamps. Planning completion does not mean every answer has finished: use list_runs for the computed status and progress.

Example prompts

OAuth scope

Requires: read:prompts

get_credits

Get the credit balance of a team, its recent use and its runway. Requires the Manager or Owner role in that team.

Parameters

Returns

OAuth scope

Requires: read:brands