generated: '2026-09-05'
method: derived
status: candidate
source: >-
  Searched 2026-09-05 for a first-party MCP server: the celery GitHub organization repository list
  (no MCP repo among 26 repos), the npm registry (celery-node, celery-ts and celery-shoot are
  independent third-party Node clients; no @celery/* MCP package), PyPI (celery-mcp -> 404), and a
  GitHub repository search. Tool candidates below are derived from the documented remote-control and
  inspect command surface at https://docs.celeryq.dev/en/stable/userguide/monitoring.html and
  https://docs.celeryq.dev/en/stable/reference/cli.html.
deployment:
  mode: none
  endpoint: null
  install: null
  package: null
  auth: unknown
  verified: searched
note: >-
  Celery publishes NO MCP server, hosted or local. This file is a CANDIDATE derivation only, and no
  `MCPServer` pointer is emitted for it — the pointer would assert an agent surface that does not
  exist. The candidate tools are one-to-one with real, documented `app.control` / `app.control.inspect`
  remote-control commands, which are the closest thing Celery has to a callable operations API.
  Note also that the remote-control transport is limited: the documentation states these commands are
  supported only on the amqp and redis brokers.
  A third-party server exists (github.com/Darius1223/celery-flower-mcp, which wraps Flower's HTTP
  API) but it is neither Celery-published nor Flower-published and is recorded here only so a later
  pass does not mistake it for first-party.
candidate_tools:
  - name: celery_status
    description: List active worker nodes in the cluster.
    backing: celery -A proj status / app.control.ping
  - name: celery_inspect_active
    description: List tasks currently being executed.
    backing: app.control.inspect().active()
  - name: celery_inspect_scheduled
    description: List scheduled ETA/countdown tasks reserved by workers.
    backing: app.control.inspect().scheduled()
  - name: celery_inspect_reserved
    description: List prefetched tasks waiting to execute.
    backing: app.control.inspect().reserved()
  - name: celery_inspect_revoked
    description: List the history of revoked task ids.
    backing: app.control.inspect().revoked()
  - name: celery_inspect_registered
    description: List task names registered on each worker.
    backing: app.control.inspect().registered()
  - name: celery_inspect_stats
    description: Return worker statistics (pool, prefetch, rusage, totals).
    backing: app.control.inspect().stats()
  - name: celery_query_task
    description: Look up the status and information for one or more task ids.
    backing: app.control.inspect().query_task(*ids)
  - name: celery_result
    description: Fetch the result and state of a task from the result backend.
    backing: celery -A proj result <task-id> / AsyncResult
  - name: celery_call_task
    description: Call a registered task by name with args/kwargs and optional eta, countdown or queue.
    backing: celery -A proj call <name>
    consequence: write — enqueues real work; Celery has no idempotency mechanism (see conventions/).
  - name: celery_revoke
    description: Revoke a task by id, optionally terminating a running task's process.
    backing: app.control.revoke(task_id, terminate=..., signal=...)
    consequence: write
  - name: celery_rate_limit
    description: Change the rate limit for a named task at runtime.
    backing: app.control.rate_limit(task_name, rate)
    consequence: write
  - name: celery_time_limit
    description: Change the soft and hard time limits for a named task at runtime.
    backing: app.control.time_limit(task_name, soft, hard)
    consequence: write
  - name: celery_pool_grow
    description: Grow the worker pool by N processes.
    backing: app.control.pool_grow(n)
    consequence: write
  - name: celery_pool_shrink
    description: Shrink the worker pool by N processes.
    backing: app.control.pool_shrink(n)
    consequence: write
  - name: celery_add_consumer
    description: Tell a worker to start consuming from a queue.
    backing: app.control.add_consumer(queue)
    consequence: write
  - name: celery_cancel_consumer
    description: Tell a worker to stop consuming from a queue.
    backing: app.control.cancel_consumer(queue)
    consequence: write
  - name: celery_purge
    description: Purge all messages from the configured task queues.
    backing: celery -A proj purge
    consequence: destructive — the documentation warns there is no undo and messages are permanently deleted.
  - name: celery_enable_events
    description: Turn the worker event stream on.
    backing: app.control.enable_events()
    consequence: write
tool_count: 19