Useful Natural‑Language Prompts for Kafka‑MCP‑Server

Below are categories of prompts you can support in a Kafka‑MCP‑Server. Each maps to one or more MCP “tools” (e.g., list_brokers, describe_topic, cluster_overview) and lets users or LLM agents interact with Kafka via plain language.

1. Cluster Metadata & Health Checks

• “List all brokers with their hostnames and ports.”

• “Show me an overview of cluster health: under‑replicated and offline partitions, controller status, and broker count.”

• “Which topics are under‑replicated right now?”

2. Topic Queries & Management

• “List all topics with partition counts and replication factors.”

• “Describe topic <topic>: show leader, replicas, and ISR per partition.”

• “Create a new topic named <name> with <n> partitions and replication factor <r>.”

• “Increase partitions on <topic> to <n>.”

• “Update <topic> config: set min.insync.replicas to <value>.”

3. Consumer Group Operations

• “List all consumer groups and their members.”

• “Describe consumer group <group-id>: show current offsets, end offsets, and lag per partition.”

• “Reset offsets for group <group-id> on <topic> to <earliest|latest|<offset>>.”

4. Message Production & Consumption

• “Publish this message to <topic>: <message-payload>.”

• “Fetch the last <n> messages from <topic>.”

• “Consume messages from <topic> starting at offset <offset> for <count> records.”

5. Security & Access Control

• “List all ACLs configured for this cluster.”

• “Show ACLs for resource <resource>, principal <principal>, and host <host>.”

• “Add an ACL to allow <principal> to <operation> on topic <topic>.”

• “Remove the ACL allowing <principal> to <operation> on <resource>.”

6. Metrics & Monitoring

• “Get throughput and latency metrics for <topic> over the last <duration>.”

• “Show broker CPU and memory usage for broker <broker-id> in the past <duration>.”

• “Alert if any consumer group lag exceeds <threshold> messages.”

7. Custom Domain Tools

• “Publish customer feedback: <text>.”

• “Retrieve sales metrics for product <product-id>.”

• “Summarize today’s error logs from the broker.”

By mapping these prompts to MCP tool invocations—such as resources/read, tools/invoke with list_topics, or cluster_overview—your Kafka‑MCP‑Server enables intuitive, conversational control and monitoring of Kafka clusters without requiring users to write code.