MCP (Model Context Protocol) server that lets AI assistants manage Confluent Cloud resources (Kafka, Connect, Flink, Tableflow, Schema Registry) through natural-language tools.
https://github.com/confluentinc/mcp-confluentStop switching between Claude and the Confluent Cloud console. This MCP server transforms your AI assistant into a Kafka operations expert, handling everything from topic management to Flink SQL through simple conversation.
If you're managing Kafka in Confluent Cloud, you know the drill: construct the right CLI command, remember the exact flag syntax, copy-paste cluster IDs, navigate between different UIs for connectors vs topics vs schemas. It's death by a thousand context switches.
This MCP server eliminates that friction entirely. Ask Claude to "show me all messages from user-events with the schema" and it handles the topic listing, schema registry lookup, and message consumption automatically. Need to create a connector? Just describe what you want - no YAML config files required.
Natural Language Kafka Operations
Intelligent Resource Management
Production-Ready Controls
Debugging Data Pipeline Issues
You: "Show me the last 10 messages from payment-events, I think there's a schema issue"
Claude: *lists messages with decoded schemas, highlights any serialization problems*
Setting Up New Data Flows
You: "Create a topic called user-analytics with 3 partitions and 7-day retention, then show me a sample producer config"
Claude: *creates topic, validates settings, provides working producer code*
Managing Stream Processing
You: "Create a Flink job that aggregates user clicks by hour from the clickstream topic"
Claude: *writes Flink SQL, submits statement, monitors execution status*
Get running in under 2 minutes:
# Quick start with npx (no installation required)
npx -y @confluentinc/mcp-confluent -e /path/to/.env
# Or clone and build from source
git clone https://github.com/confluentinc/mcp-confluent.git
cd mcp-confluent
npm install && npm run build
Configure your AI client to connect (detailed setup for Claude Desktop and Goose CLI included in the repo).
Create a .env file with your Confluent Cloud credentials. The server supports the full Confluent ecosystem:
The CLI provides precise control over which operations are available:
# Enable only safe read operations
npx @confluentinc/mcp-confluent --allow-tools list-topics,consume-messages,list-schemas
# Block destructive operations
npx @confluentinc/mcp-confluent --block-tools delete-topics,delete-connector
# Multiple transports for different use cases
npx @confluentinc/mcp-confluent --transport http,sse,stdio
Perfect for production environments where you want AI assistance without risking accidental deletions.
Managing modern data infrastructure shouldn't require memorizing dozens of CLI commands or navigating multiple web consoles. Your AI assistant already understands your intent - now it can execute it directly against your Kafka infrastructure.
This isn't just convenience; it's a fundamental workflow improvement. Faster debugging, easier experimentation, and lower cognitive load when working with streaming data systems.
Built and maintained by Confluent Inc., so you're getting official support for the tools you already use in production.
Get started: npx -y @confluentinc/mcp-confluent --help