v0.5.1 — @getworkbench/mcp
✨ New package: `@getworkbench/mcp`
Optional Model Context Protocol server that lets Cursor, Claude Desktop, Zed, Continue.dev, and any other MCP-aware agent inspect, debug, and operate BullMQ queues through a running Workbench dashboard.
Why?
Workbench gives developers a beautiful BullMQ dashboard. The MCP gives agents the same surface. Same data, same operations, same permissions — but reachable from an AI editor's chat:
- "Why is `email-send` backed up?" — agent calls `workbench_get_metrics`, sees p95 latency spiked at 14:02, calls `workbench_list_runs` with `status=failed` and `from=14:00`, finds the regression.
- "Retry every failed `webhook-deliver` from the last hour." — `workbench_list_runs` + `workbench_bulk_retry`.
- "Run the nightly billing scheduler now so I can verify the fix." — `workbench_list_schedulers` + `workbench_run_scheduler_now` (added in 0.5.0).
18 tools
Inspect (`readOnlyHint: true`): `workbench_get_overview`, `list_queues`, `get_quick_counts`, `get_metrics`, `get_activity`, `list_jobs`, `list_runs`, `get_job`, `search_jobs`, `list_schedulers`, `list_flows`, `get_flow`, `list_tag_values`
Operate (`destructiveHint: true`): `workbench_retry_job`, `remove_job`, `promote_job`, `pause_queue`, `resume_queue`, `run_scheduler_now`, `enqueue_job`, `clean_jobs`, `bulk_retry`, `bulk_delete`
Install
```json
{
"mcpServers": {
"workbench": {
"command": "npx",
"args": ["-y", "@getworkbench/mcp"],
"env": {
"WORKBENCH_URL": "http://localhost:3000/jobs",
"WORKBENCH_USERNAME": "admin",
"WORKBENCH_PASSWORD": "hunter2"
}
}
}
}
```
The MCP is a thin HTTP proxy to the dashboard's REST API — it reuses the dashboard's existing Basic Auth and inherits the `readonly` flag end-to-end. No second source of truth for permissions.
Full docs: packages/mcp/README.md
Released packages
- `@getworkbench/mcp@0.5.1` (new)
All other packages remain at `0.5.0`.