Nº 02 — Open-source BullMQ dashboard
Better Bull Board
A modern BullMQ dashboard: real-time job streams, logs, metrics, and an MCP server so coding agents can query the queues.
Nº 02 — Diagram
- Year
- 2025
- Role
- Core contributor
- Stack
- Next.jsTypeScriptBullMQRedis StreamsPostgreSQLClickHouseDrizzleMCPKubernetes
01Context
Classic BullMQ dashboards show the queues as they are now, not their history: there is no telling why a job failed yesterday, or which queue has been backed up for an hour.
02What I built
- An ingester that consumes job events through Redis Streams consumer groups, with periodic reconciliation against BullMQ's state.
- A Next.js interface: queues, schedulers, runs and logs, with metrics in ClickHouse.
- An MCP server: an agent can ask which queues are backed up or why jobs fail, and, with write access, pause, replay or cancel.
- A client published on npm,
@better-bull-board/client, versioned by semantic-release.
03Technical decisions
- Polling for long-lived data (queues, schedulers), ingestion for volume (jobs, logs): no need to wrap every queue with the client.
- The app and the ingester deploy separately on Kubernetes, and both scale horizontally.
- A 30-day retention on runs and logs.