All Posts

We pointed one proposal engine at 1,000 randomly generated Kafka clusters - three brokers to a thousand, a thousand partitions to a million, every rack and replication shape we could think of. No per-cluster tuning. Here is how little the answer moves when you move everything else.

Just shutting down a broker that's getting replaced means hours of running with thousands of under-replicated partitions. Pilot's maintenance endpoint drains the broker first - one API call, no manual replica planning.

Pilot started as a metadata-only Kafka tool. The new Pilot Agent is opt-in, lives next to your broker, and unlocks what the protocol cannot show.

How Pilot's MCP-powered AI agent traced slow order processing to a disk-hot broker, rebalanced the affected topic, and resolved consumer lag - all through natural language with human approval gates.

We've published comprehensive documentation for Calinora Pilot at docs.calinora.io - getting started guides, feature deep-dives, deployment recipes, and a full API reference.

Every Kafka team hits the same wall. Broker load drifts, partitions need moving, and figuring out what's safe takes too long. That's why we built Pilot.

Introducing the Calinora blog - your source for product updates, Kafka best practices, and technical deep-dives.