list_workersLABSCO SUMMARY
Diagnose a backed-up queue and unblock it in the same conversation.
Both halves are here: list the workers and the queue depths to find the jam, then grow the pool, attach a consumer or revoke the task that is wedging it. Worth knowing before you connect it that the same surface reaches as far as shutting a worker down.
What it is
A wrapper over the Celery Flower REST API that exposes its endpoints as MCP tools. Point it at a running Flower instance and worker pools, queues and tasks become things an assistant can inspect and act on.
What you get
- Every worker and its state in real time (list_workers)
- Pool control: grow, shrink, autoscale, restart or shut a worker down (grow_worker_pool, shrink_worker_pool, autoscale_worker_pool, restart_worker_pool, shutdown_worker)
- A queue consumer added on a running worker (add_queue_consumer)
- Task management: apply, revoke, abort, and set timeouts and rate limits
- Queue inspection, including depths and consumers
- Full coverage of the Flower REST API, not a chosen subset
Requirements
Python and a reachable Flower instance: FLOWER_URL=http://localhost:5555 uvx celery-flower-mcp. Anything the Flower API can do to your cluster these tools can do, shutdown_worker included.
Setup effort
One command — FLOWER_URL=http://localhost:5555 uvx celery-flower-mcp
USE CASES
01
Working out why a queue is not draining
Read the workers and their states next to the queue depths and consumer lists. Usually the answer is that nothing is consuming the queue, which is visible in one pass rather than three dashboards.
02
Absorbing a burst without redeploying
Grow the pool on the workers already running, or hand it to autoscale with bounds, and shrink it back afterwards. No config change, no restart of the service.
03
Getting a stuck task out of the way
Revoke or abort the task that is holding a worker, and put a rate limit or a timeout on the task type so the same one does not do it again this afternoon.
04
Taking a worker out of service cleanly
Restart the pool to pick up new code, or shut the worker down when you are draining a host. Both are single calls, and both are irreversible from the assistant's side, so they are worth confirming first.
21 tools
Listed in project docs. Your client may show a different list.
shutdown_workerrestart_worker_poolgrow_worker_poolshrink_worker_poolautoscale_worker_pooladd_queue_consumercancel_queue_consumerlist_taskslist_task_typesget_task_infoget_task_resultapply_taskasync_apply_tasksend_taskabort_taskrevoke_taskset_task_timeoutset_task_rate_limitget_queue_lengthshealthcheckBEFORE YOU INSTALL
What to know first
Tools you get
21 tools named in the project's documentation.
Requires
uvx on PATH
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
FLOWER_URL=http://localhost:5555 uvx celery-flower-mcpMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"celery-flower": {
"command": "uvx",
"args": [
"celery-flower-mcp"
],
"env": {
"FLOWER_URL": "http://localhost:5555"
}
}
}
}
