bunqueue MCP
The MCP server of the bunqueue job queue: an agent schedules cron jobs, retries failures, sets limits and reads stats, no Redis.
- 592 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The agent can push, retry and delete jobs and register HTTP addresses that a worker calls for every job.
- Without the agent profile, destructive calls run without confirmation.
Install
- Works in
Manual install
claude mcp add bunqueue -e BUNQUEUE_MCP_PROFILE=agent -- bunx --package=bunqueue bunqueue-mcpREADME command. The @modelcontextprotocol/sdk package is an optional dependency: bun add bunqueue @modelcontextprotocol/sdk.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent can push, retry and delete jobs and register HTTP addresses that a worker calls for every job.
- Without the agent profile, destructive calls run without confirmation.
This is third-party code. Review the repository files before installing.
What it does
bunqueue is a job queue for Bun with SQLite by default and PostgreSQL for multiple brokers. It has a dead letter queue, cron, SQLite backups to S3 and an MCP server built in, and needs no Redis. This card describes the `bunqueue-mcp` server: 79 primary tools, five resources and three prompts. Through them an agent schedules cron jobs, pushes and processes jobs, retries failures, sets rate limits and reads stats without glue code. The tools include bounded paging, job observation and queue diagnosis. HTTP handlers let an agent register an address that a worker calls for every job. Two scoped resource templates limit reads to one queue, and queue pages omit job payloads by default. Up to three workflow tools can be enabled on top. The server runs over stdio or Streamable HTTP. The `agent` profile turns on dynamic toolsets and confirmation of destructive calls. Configuration can be checked before connecting with `BUNQUEUE_MCP_PROFILE=agent bunx --package=bunqueue bunqueue-mcp --doctor --json`, which does not open embedded databases.
Who it is for. Developers on Bun and Node.js whose background jobs and schedules live in bunqueue and who want an agent to manage queues and triage failures.
Good fit when
- In the morning you want every failed job listed and the ones whose cause is fixed retried.
- A daily cleanup has to be set for 3 AM with one sentence.
- A queue for an external API needs a rate limit of 50 jobs per second.
- The agent needs a diagnosis of why a queue is stuck: where jobs stall and what sits in the dead letter queue.
- Long agent steps need a queue that keeps state in SQLite instead of Redis.
Not a fit when
- The queue runs on BullMQ or another broker: the server manages only bunqueue.
- The project has no Bun: the server is launched with `bunx`.
- An agent must not manage jobs of a production queue without confirmation: the default profile does not enable it.
Example request
Show all failed jobs in the emails queue, retry the ones that failed on timeout and set a rate limit of 50 per second on the api-calls queue.MCP
- Transport
- stdio, http
- Authentication
- not required
Environment variables
BUNQUEUE_MCP_PROFILE- Server profile; the README examples set it to agent.
Limitations
Bun is required, and the `@modelcontextprotocol/sdk` package is installed as an optional dependency. Without the `agent` profile, dynamic toolsets and confirmation of destructive calls are off.
For durable agent turns the author points to `Workflow`, `Engine` and the `bunqueue/agents` journal. Session resume and queue retries still require provider idempotency and manual reconciliation of external effects. The final job state must be read before claiming success.
How to disable. Run `claude mcp remove bunqueue` or delete the bunqueue entry from your client's `mcpServers`.
FAQ
How do I connect bunqueue to Claude Code?
With `claude mcp add bunqueue -e BUNQUEUE_MCP_PROFILE=agent -- bunx --package=bunqueue bunqueue-mcp`. For Claude Desktop, Cursor and Windsurf, put the same `bunx` command into `mcpServers`.
How many tools does the server have?
79 primary tools, five resources and three prompts. Up to three workflow tools can be added, and dynamic discovery tools are counted separately.
Is Redis required?
No. Data lives in SQLite by default, and a multi-broker mode on PostgreSQL is available when you scale.
Which clients and transports are supported?
Claude Code, Claude Desktop, Cursor and Windsurf. The server runs over stdio or Streamable HTTP, and the README gives sample configurations.
How do I check the setup in advance?
Run `BUNQUEUE_MCP_PROFILE=agent bunx --package=bunqueue bunqueue-mcp --doctor --json`. It prints readiness and does not open embedded databases.
Discussion
No questions or reviews yet.
Related
MCP serversAn MCP server for Lunora apps: the agent reads functions, tables, logs and errors of a deployed backend and runs queries.
Rebase MCP server: the agent edits the schema, documents and users of a Postgres backend, with writes limited to localhost
An MCP server and plugin for agents that generate web APIs from OpenAPI, AsyncAPI, GraphQL, gRPC and SQL specs on a Rust core.
A PostgreSQL MCP server with index tuning, EXPLAIN plans, database health checks and a read-only mode
Apollo's MCP server: turns chosen GraphQL operations into tools the agent uses to work with your API
Bitrix24 MCP Server (lstkhN)
Bitrix24 MCP Server
A module for on-premise Bitrix24: an agent safely reads CRM, the MySQL schema, code and business processes over SSH stdio or HTTP with a Bearer token
README in short
The README opens with the queue overview, then API examples, architecture, features, workflows and a "Built for AI Agents" chapter with the MCP server, connection commands and sample phrases. Sections on the dashboard, SDKs, CLI and production use follow.