get_run

Use get_run to poll or inspect one run by run_id; if truncated is true, call get_run_output(run_id, output_name) directly using a name from output.outputs_index instead of re-polling or asking the user; also prefer get_run_output when only one output is needed.

read-only · idempotent

Parameters

Name	Type	Required	Default	Description
run_id	Annotated[str, Field(description='UUID of the run to look up, or a UUID prefix (hex digits and dashes only). Use list_runs to discover run IDs for a model.')]	Yes	—
workspace_id	`Annotated[str	None, Field(description=ID_BACKED_WORKSPACE_ID_FIELD_DESCRIPTION)]`	No	None

Returns

Returns a structured object. Top-level fields:

• run_id (string)
• run_number (null)
• model_name (string)
• status (string)
• commit_sha (string)
• started_at (string)
• finished_at (string)
• error_message (null)
• output (object)
• truncated (boolean)

See the response example below for the full payload shape.

Example

Tool Call

{
  "name": "get_run",
  "arguments": {
    "run_id": "66666666-6666-6666-6666-666666666666"
  }
}

Response

{
  "run_id": "66666666-6666-6666-6666-666666666666",
  "run_number": null,
  "model_name": "forecasts",
  "status": "success",
  "commit_sha": "abc1234def5678901234567890abcdef12345678",
  "started_at": "2026-04-28T10:00:00+00:00",
  "finished_at": "2026-04-28T10:00:01.420000+00:00",
  "error_message": null,
  "output": {
    "status": "success",
    "exit_code": 0,
    "stdout": "Run completed in 1.42s\n",
    "stderr": "",
    "outputs": {
      "revenue": [
        100,
        110,
        121
      ]
    }
  },
  "truncated": false
}

Handling a truncated response

When the persisted run payload exceeds the inline cap, output is replaced with a compact index instead of the full payload:

{
  "run_id": "66666666-6666-6666-6666-666666666666",
  "status": "success",
  "output": {
    "truncated_payload": true,
    "outputs_index": ["revenue", "expenses", "headcount"],
    "outputs_count": 3,
    "total_bytes": 184320,
    "hint": "Payload exceeds 65536 byte get_run inline cap. Use get_run_output(run_id, output_name) to fetch individual outputs by name."
  },
  "truncated": true
}

Do not re-call get_run or ask the user which output to fetch. Pick a name directly from output.outputs_index and call get_run_output with it — that is the entire recovery path.

Notes

• Module: Execution
• Requires authentication via API token or Auth0 JWT.