elastic/mcp-server-elasticsearch — reverse-engineered prompt
Reverse engineered prompt
Build me a small server that lets AI tools like Claude Desktop, Cursor, or VS Code talk to an Elasticsearch cluster through MCP. I want it to run in Docker and support both direct stdio mode for local clients and an HTTP mode for web style connections or multiple users.
It should read settings from environment variables for the Elasticsearch URL, either an API key or username and password, and an optional setting to skip SSL checks for local testing. Add a simple health endpoint that returns pong in HTTP mode, and make the startup errors clear when credentials or the cluster URL are wrong.
The goal is that once it’s running, an MCP client can connect and ask questions about Elasticsearch indices, retrieve data, and analyze results without me writing custom API calls. Include a basic example config for Claude Desktop, a sample env file, Docker instructions, and a short troubleshooting section. Look up the current MCP docs online if you need to.
Want more depth? Deep Reverse