Deploy KubeMQ
One broker that replaces your messaging stack and runs your AI-agent fabric — pick your path and install in minutes.
One broker, two jobs
KubeMQ is a Kubernetes-native message broker that ships as a single binary with two co-equal headline capabilities.
Consolidate your messaging stack — point your existing Kafka, RabbitMQ, AWS SQS/SNS, MQTT, STOMP, AMQP, or Google Cloud Pub/Sub client at KubeMQ by changing one connection string. No rewrite, no new SDK, no new wire protocol to learn.
Run your AI-agent fabric — KubeMQ ships built-in Agent-to-Agent (A2A) and Model Context Protocol (MCP) gateways. Agents register, get discovered by capability, and stream results; any MCP-compatible LLM host — Claude, Cursor, or your own IDE — drives both messaging and agents over the open MCP standard.
Both capabilities ride the same core:
- 10-language SDKs plus REST
- Four messaging patterns (Events, Events Store, Queues, and RPC — Commands and Queries)
- Built-in persistence
- A web dashboard
- Clustering and high availability
- No ZooKeeper, no external database, no sidecars
Choose your path
What do you want to do
Two things KubeMQ does — pick the one you came for.
Replace your messaging stack
Point your existing Kafka, RabbitMQ, AWS SQS/SNS, MQTT, STOMP, AMQP, or Google Cloud Pub/Sub client at KubeMQ — change one connection string, keep your code.
Build on the AI-agent fabric
Connect an MCP host, register an agent, or drive messaging and agents from an LLM over the built-in Agent-to-Agent (A2A) and Model Context Protocol (MCP) gateways.
Ways to get started
Or just get it running:
Build native (SDKs)
Run KubeMQ, open the dashboard, and round-trip your first message using a native SDK.
Drive from the terminal (kmq CLI)
Install the kmq CLI, connect to a running server, and send your first message without writing code.
Deploy to production
Install KubeMQ on Kubernetes with the Helm charts and operator, with a production readiness checklist.
The four messaging patterns
Events
Fire-and-forget pub/sub for real-time broadcasting to multiple subscribers.
Events Store
Persistent pub/sub with replay, retention policies, and consumer groups.
Queues
Point-to-point messaging with guaranteed delivery, acknowledgment, and dead-letter queues.
RPC
Request-reply messaging for commands (execute and confirm) and queries (request data).
Upgrading an existing install?
Coming from KubeMQ v2? Follow the v2 → v3 migration guide. Rolling upgrades, rollback, and backup are covered in Install with Docker and Install on Kubernetes.
Was this page helpful?
KubeMQ Documentation
Kubernetes-native enterprise message broker. Connect existing Kafka, RabbitMQ, MQTT, SQS, and Pub/Sub clients unchanged — or build with SDKs in 10 languages.
Quickstart
Run KubeMQ, open the dashboard, and round-trip your first message from the terminal — the ~5-minute universal win.