Install guide and config at curatedmcp.com

Upstash MCP: Give Claude Direct Access to Serverless Redis and Kafka

Managing caches, rate limits, and message queues shouldn't require context-switching out of your AI agent. Upstash MCP bridges that gap by giving Claude, Cursor, and Windsurf direct access to Upstash's serverless Redis and Kafka services — all through a single, unified interface.

What It Does

Upstash MCP unlocks two core capabilities for AI agents:

Redis operations: Get, set, and delete keys with full TTL support. Work with strings, lists, hashes, sets, and sorted sets. Inspect key patterns, check database statistics, and manage multiple databases across regions — all without leaving your agent's context.

Kafka management: Create and manage topics, inspect consumer group lag, and monitor message throughput directly from Claude.

This is particularly valuable if you're already using Context7 MCP (Upstash's documentation layer) — now you can manage the Redis cache powering it directly from your agent. The pay-per-request pricing model means zero idle costs, making it perfect for intermittent AI workloads that don't justify always-on infrastructure.

How to Install

Install via npm:

npx -y @upstash/mcp-server

Add to your Claude Desktop claude_desktop_config.json:

{
  "mcpServers": {
    "upstash-mcp": {
      "command": "npx -y @upstash/mcp-server",
      "env": {
        "UPSTASH_REDIS_REST_URL": "your-redis-url",
        "UPSTASH_REDIS_REST_TOKEN": "your-redis-token",
        "UPSTASH_KAFKA_REST_URL": "your-kafka-url",
        "UPSTASH_KAFKA_REST_TOKEN": "your-kafka-token"
      }
    }
  }
}

Get credentials from your Upstash console.

Real-World Use Cases

• Cache inspection during debugging: Ask Claude to check what's currently cached, inspect TTLs, and clear specific keys without SSH-ing into infrastructure.
• Rate limit management: Have your agent adjust rate-limit counters in Redis or check current thresholds while troubleshooting API behavior.
• Event queue monitoring: Query Kafka consumer lag, topic throughput, and message rates to diagnose streaming pipeline issues in real time.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.