🌊 Agentic Engineering for Apache Kafka

A collection of agent skills that turn AI agents and coding tools such as Claude Code and Cursor into Kafka-specialised engineering assistants. Audit topic configurations, diagnose consumer lag, review schema changes, review connectors and DLQs, catch security misconfigurations and tune performance, all from a single prompt instead of 15 minutes of manual exploration or investigation.

Maintained by Lenses.io, the team that pioneered the developer experience for Apache Kafka. Agentic engineering has shifted what that means. Making sure an AI agent knows how to handle streaming data is now part of the job.

Why Skills and MCP?

A Kafka MCP server gives agents access to your live cluster: topics, consumer groups, connectors, schemas and metrics. The skills in this repository teach agents expertise: best practices, audit thresholds, remediation playbooks and standard workflows. Together they enable AI-powered Kafka engineering where the agent doesn't just read your cluster, it knows what to look for and how to fix it.

Without skills, agents are confident generalists. They will write a consumer for the orders topic that compiles and runs but does not handle deserialization errors properly, set up DLQs correctly, or partition consumption sensibly for the topic's layout. Skills close the gap between code that runs in a demo and code that holds up in production.

These skills follow the Anthropic open standard for skills, so they are portable across Claude Code, Cursor, Claude.ai and the Claude Messages API. They are MCP-agnostic by design: every skill in this repo is tested against Lenses MCP Server (the recommended setup), but will work with any Kafka MCP server that exposes similar tools.

The quickest way to try the skills end-to-end is with the free Lenses Community Edition, which ships with Lenses HQ, a remote MCP Server and a pre-configured single-broker Kafka cluster with demo data.

What's Included

Kafka skills