Version
ASYNC

get_job_status

Poll a long-running MCP job by id. Returns the current status plus result or error once the worker finishes; in the Copilot the step follows the job live instead.

Poll a long-running MCP job by id. Returns the current status plus result or error once the worker finishes.

Status values: queued → running → completed | failed.

Pair it with any tool that hands back a job_id: bulk_generate_access_codes, request_account_export, and revert_recovery_operation for a channel revert, whose members are re-admitted in the background.

In the Copilot none of this polling is needed: the step under the reply follows the job live, "In progress" until it ends with "Done, 500 items" or the error, because every mark the worker takes reaches the creator's feed. A broadcast is not a job: it has a record of its own, read with get_broadcast.

Annotations

Read-only

It reads and never changes anything.

Arguments

job_id*string

UUID of the async job to poll.

What it returns

{  "data": {    "job_id": "0a4e7b96-c358-4d12-9f6b-25a8013ce74f",    "tool_name": "bulk_generate_access_codes",    "status": "completed",    "result": { "count": 50, "batch_unique_key": "..." },    "error": null,    "started_at": "2026-05-18T10:05:00Z",    "completed_at": "2026-05-18T10:05:04Z"  }}

result and error are both null until the worker finishes; exactly one is populated afterwards. started_at is null while the job is still queued. result.count is what the worker counted: codes minted, or members queued for re-admission; an export carries its download instead.

How it fails

AUTHENTICATION_REQUIRED

no authenticated user on the request.

RESOURCE_NOT_FOUND

unknown job_id, or the job was started by a different user.

How is this guide?

Last updated on