Create an automation run

Request

Create a pending run with a free-text run prompt and a snapshot of the saved automation instructions. Any member of the automation's project, or an organization API key, can start a run; it executes as the automation's owner, using the owner's data access and AI usage.

Security

ApiKey(Required scopes: automations.run)

Bodyapplication/jsonrequired

automationIdstring, non-empty

The automation to run. Required.

runPromptstring, non-empty\S

Free-text instructions for this run. May reference recordings or other resources. Required for members, including personal access tokens. Organization API keys must omit it, and the run uses only the automation's saved instructions.

Payload

{
  "automationId": "string",
  "runPrompt": "string"
}

Responses

The pending run

Bodyapplication/json

idstringrequired

automationIdstringrequired

actionPromptstringrequired

Snapshot of the automation's saved instructions.

runPromptstringrequired

Prompt submitted for this run.

statusstringrequired

Enum:"pending""running""succeeded""failed"

createdAtstring, (date-time)required

updatedAtstring, (date-time)required

startedAtstring or null, (date-time)required

completedAtstring or null, (date-time)required

errorMessagestring or nullrequired

executionstringrequired

Whether Foxglove ran the automation to the end, separate from what the run found. finished means the agent's turn ended, or a limit stopped the run after the agent reported its result. blocked means a limit or a missing permission stopped the run before that, such as the AI usage limit, the run's time limit, or an owner who lost access to the project. error means the run failed on Foxglove's side.

Enum:"pending""running""finished""error""blocked"

resultstring or nullrequired

What the agent reported the run accomplished. needs_review means the task finished but the agent was not sure whether something went wrong. no_result means the task ran and had nothing to act on, such as no matching data. Null when the agent did not report a result, for example because the run was blocked or failed before it finished.

Enum:"succeeded""needs_review""failed""no_result"

resultSummarystring or nullrequired

The agent's one-sentence summary of what the run found or did.

modelstring or nullrequired

Concrete model the run's chat used, such as claude-sonnet-5-5. Null until the run's chat exists.

Response

{
  "id": "string",
  "automationId": "string",
  "actionPrompt": "string",
  "runPrompt": "string",
  "status": "pending",
  "createdAt": "2019-08-24T14:15:22Z",
  "updatedAt": "2019-08-24T14:15:22Z",
  "startedAt": "2019-08-24T14:15:22Z",
  "completedAt": "2019-08-24T14:15:22Z",
  "errorMessage": "string",
  "execution": "pending",
  "result": "succeeded",
  "resultSummary": "string",
  "model": "string"
}