Skip to main content
View Criterion Runs, the per-criterion results produced by the verification engine.
The MCP server currently still exposes the legacy eval-runs tool. The payload shape is equivalent (passed, score, details), except a Criterion Run includes structured evidence. A dedicated criterionRuns tool is planned.

​
Actions (planned)

ActionDescription
listList all criterion runs for a task run
getGet a criterion run by ID

​
Parameters

​
list

ParameterTypeRequiredDescription
taskRunIdstringyesTask run ID to list criterion runs for

​
get

ParameterTypeRequiredDescription
idstringyesCriterion run ID

​
Examples

​
List criterion runs for a task run

{
  "action": "list",
  "taskRunId": "tr_abc123"
}

​
Get a criterion run with evidence

{
  "action": "get",
  "id": "crun_01H..."
}