queue-pilot

Queue Pilot is an MCP server that bridges your AI assistant with message brokers (RabbitMQ and Apache Kafka) to inspect, validate, and manage messages using JSON Schemas. It exposes a unified set of universal tools (like list_schemas, inspect_queue, validate_message, publish_message, purge_queue, and more) that let you peek at queue contents, validate payloads against schemas, and even publish validated messages back to brokers. This enables safer, schema-driven development and quick debugging of message flows across multiple brokers from a single interface. To start, configure the server with your schemas and connect it to your broker(s); then use your MCP client to issue tool commands (such as list_queues, peek_messages, inspect_queue, or validate_message) to explore and validate your message workloads. The server supports both RabbitMQ and Kafka, with broker-specific capabilities like listing exchanges and bindings (RabbitMQ) or consumer groups and partitions (Kafka).

Prerequisites:\n- Node.js version >= 22 (check with node --version)\n- npm (comes with Node.js)\n- Access to a message broker (RabbitMQ with management plugin enabled, or Apache Kafka with the kafka-javascript peer dependency)\n\nInstallation steps:\n1) Ensure Node.js is installed. Verify: node --version\n2) Use npm/npx to run Queue Pilot directly from the registry without a global install (recommended for MCP usage):\n\nbash\nnpx -y queue-pilot init --schemas /absolute/path/to/your/schemas --client <name>\n\nThis will generate an MCP config for your client. If you prefer manual configuration, you can add a server entry like:\n\njson\n{\n "mcpServers": {\n "queue-pilot": {\n "command": "npx",\n "args": [\n "-y",\n "queue-pilot"\n ]\n }\n }\n}\n\n3) If you are developing locally from source, you can run the server via npx as shown above or use tsx to run the TypeScript source, for example:\n\nbash\nnpx queue-pilot init --schemas ./schemas --client claude-code\nnpx tsx src/index.ts --schemas ./schemas\n\n4) For Kafka integration, ensure you have the required peer dependency and provide broker information via environment variables or config as needed.\n5) When configuring the MCP client, place the resulting JSON config in your typical MCP config path (as described in the README for your client).

Tips and notes:\n- Use absolute paths for the --schemas option when generating or using manual configurations to avoid path resolution issues.\n- If npx fails to resolve the package on Windows, try cmd /c npx queue-pilot init ....\n- For Kafka, you may need to include the Kafka JavaScript client as a peer dependency and configure KAFKA_BROKERS in env vars.\n- Secrets (like broker credentials) are often injected via environment variables (not command output) to reduce exposure in process listings.\n- The MCP config supports both RabbitMQ and Kafka; use the universal tools to explore schemas, queues/topics, and messages, then switch to broker-specific tools as needed.\n- When running from source, you can point the server to your local schemas directory with --schemas ./schemas and set the working directory appropriately (cwd) if required by your MCP client.