Delv
Official (Vendor)Active· 11d4.3by Amazon MQ

RabbitMQ MCP

Interacts with RabbitMQ brokers for admin operations and message enqueue or dequeue, with OAuth and HTTP transport.

A
Safety & Trust

Delv Safety Grade: A

Score 84/100 · assessed 2026-04-28

Maintainer95
Permissions65
Supply chain85
Transparency80
Incidents100

Amazon MQ's official RabbitMQ MCP server is backed by a major cloud vendor with strong infrastructure credentials. The maintainer score is excellent given AWS's resources and commitment to enterprise tooling. Permissions are moderately scoped: it can read and write messages, manage queue topology, and perform admin operations on your broker, which is appropriate for its debugging and operations use case but does grant substantial control over your messaging infrastructure. Supply chain is solid via PyPI package distribution, though the repository is relatively new. Transparency is good with open source code and clear documentation of capabilities. No known security incidents. The main risk is the breadth of RabbitMQ admin permissions combined with credential access, so treat the RABBITMQ_URL environment variable as highly sensitive.

Lethal Trifecta (prompt-injection exposure)

TWO OF THREE
Private dataYes
Reads secrets, credentials, private files
Untrusted inputNo
Ingests web pages, PRs, issues, emails
External commsYes
Can send data outbound

Queue contents may include private payloads; publishing is outbound. No web-style input ingestion.

Green flags

  • Official Amazon MQ vendor implementation with enterprise backing
  • Uses RabbitMQ Management API over HTTP rather than raw AMQP protocol
  • OAuth-aware authentication support for modern security patterns
  • Distributed via standard PyPI package registry with versioning
  • Clear documentation of admin vs read-only capabilities

Red flags

  • Requires full broker credentials via RABBITMQ_URL environment variable
  • Can delete queues and exchanges, potentially disrupting production systems
  • Mutative tools flag indicates write operations enabled by default
  • Network access to potentially internal message broker infrastructure

Permissions requested

Outbound networkPrivate networkAccess secretsSend messagesRead messagesDB readDB write
Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Install

uvx amq-mcp-server-rabbitmq@latest --allow-mutative-tools
Env vars needed: RABBITMQ_URL

Review

This is Amazon MQ's official MCP server for RabbitMQ, and it does exactly what the name suggests: lets Claude talk directly to your RabbitMQ broker. You can inspect exchanges, queues, bindings, publish messages, consume from queues, and run basic admin tasks without leaving your editor. It's OAuth-aware and uses HTTP transport, which means it plays nicely with RabbitMQ's management API rather than trying to speak AMQP directly. I'd reach for this when debugging message flows in a distributed system. Say you've got a dead-letter queue piling up and you need to peek at the messages, understand why they're failing, and maybe replay a few. Normally that's a browser tab, some curl commands, and a lot of copy-pasting. With this MCP, you just ask Claude to show you the messages, explain the error patterns, and republish the fixed ones. It's surprisingly fluid once you've got it wired up. The mutative tools flag is worth noting. By default, this server can publish and delete messages, which is powerful but also means Claude can accidentally nuke a queue if you phrase a request badly. The flag makes that explicit, which I appreciate. The OAuth support is solid if you're running RabbitMQ in a managed environment like Amazon MQ, though local setups with basic auth work fine too. Quirks: it's Python-based and uses uvx, so you need a working Python environment. The config is straightforward but you do need to know your broker's management API URL upfront. If you're running RabbitMQ locally, that's usually http://localhost:15672. If you're on a managed service, you'll need the full endpoint. No autodiscovery here. Who shouldn't bother: if you're not actively working with RabbitMQ, this is useless. If you only occasionally check queues and you're happy with the web UI, the overhead isn't worth it. But if you're debugging message flows daily, or if you're building tooling around RabbitMQ and want to prototype admin workflows in natural language, this is quietly excellent. It's one of those MCPs that doesn't try to do too much and nails what it does attempt.
Verdict

Install this if you work with RabbitMQ regularly and want to debug or administer queues without context-switching to a browser. Skip it if you only touch RabbitMQ occasionally or if you're happy with the management UI. It's a focused tool that does one thing well.

Good at

  • Lets you inspect queues, peek at messages, and replay dead-lettered messages without leaving your editor.
  • OAuth support makes it work cleanly with managed RabbitMQ services like Amazon MQ.
  • Mutative tools flag is explicit, so you know when Claude can modify state.
  • Uses the management API over HTTP, which means no AMQP protocol complexity.
  • Official vendor support from Amazon MQ, so it's maintained and documented properly.

Watch out

  • Requires a working Python environment and uvx, which adds setup friction on some machines.
  • No autodiscovery of brokers, so you need to know your management API URL upfront.
  • Mutative operations can be dangerous if you phrase requests ambiguously.
  • Only useful if you're already working with RabbitMQ daily.

Use cases

  • broker inspection
  • queue administration
  • message replay
  • debugging distributed systems

Getting started

1. Run `uvx amq-mcp-server-rabbitmq@latest --allow-mutative-tools` to install. The mutative flag lets Claude publish and delete messages. 2. Add the server to your Claude Desktop config with `RABBITMQ_URL` set to your broker's management API endpoint (e.g., http://localhost:15672 for local, or your Amazon MQ HTTPS endpoint). 3. Restart Claude Desktop and ask it to list your RabbitMQ queues. If it returns queue names, you're connected. 4. Watch out for the mutative tools. Claude can delete messages or purge queues if you ask it to, so phrase requests carefully until you're comfortable with the behaviour.

Works with

Claude DesktopClaude CodeCursor

Similar MCPs