dockndevai/mcp-kafka

A Model Context Protocol server for **Apache Kafka**. It lets an MCP-capable client (Claude Desktop, Claude Code, etc.) **monitor and manage** Kafka clusters — topics, partitions, configs, and consumer groups (including lag) — with behaviour controlled entirely by flags.
Safe by default: it starts read-only, can be scoped to an allowlist of topics, protects internal/critical topics from mutation, and gates destructive operations behind an explicit opt-in.
Features
- **Monitoring** — cluster/broker info, topic metadata and offsets, consumer groups, and **per-partition + total consumer lag**.
- **Management** — create topics, add partitions, alter topic configs, reset group offsets; delete topics/groups (admin).
- **Access modes** — read-only → read-write → admin, layered so a mode never exposes tools above its level.
- **Security flags** — topic allowlist, protected/internal topics, delete gating, dry-run, and JSON audit logging (see below).
- **Auth** — plaintext, TLS, and SASL (PLAIN / SCRAM-SHA-256 / SCRAM-SHA-512).
Security model
| Concern | Flag | Default | Effect |
| --- | --- | --- | --- |
| What can the server do? | KAFKA_MODE | read-only | read-only exposes only monitoring; read-write adds management; admin adds deletes. Tools above the mode are **never registered**. |
| Which topics are in scope? | KAFKA_TOPIC_ALLOWLIST | *(all)* | When set, operations on other topics are refused. |
| Protect internal topics | KAFKA_PROTECT_INTERNAL_TOPICS | true | Topics starting with _ can be read but never mutated. |
| Protect specific topics | KAFKA_PROTECTED_TOPICS | *(none)* | Additional read-only-forever topics. |
| Can it delete? | KAFKA_ALLOW_DELETE | false | delete_topic / delete_consumer_group need this **and** admin mode. |
| Preview without touching the cluster | KAFKA_DRY_RUN | false | Write/admin tools validate + log intent, then return. |
| Audit trail | KAFKA_AUDIT_LOG | true | Emits a JSON line to stderr per guarded operation. |
| Interactive confirmation | *(automatic)* | — | Destructive & high-impact actions prompt the human to approve via MCP elicitation before running; clients without elicitation fall back to the *_ALLOW_* gate. |
Tools
**Read** (read-only+): cluster_info, list_topics, describe_topic, topic_offsets, list_consumer_groups, describe_consumer_group (with lag)
**Write** (read-write+): create_topic, create_partitions, alter_topic_config, reset_consumer_group_offsets
**Admin** (admin): delete_topic, delete_consumer_group (both need KAFKA_ALLOW_DELETE)
Quickstart — add to your agent
Published on npm as @dockndevai/mcp-kafka. No clone or build needed — your MCP client runs it on demand with npx. **Start in read-only mode**; see .env.example for every variable and docs/CLIENTS.md for the full per-client guide.