Elasticsearch MCP — Search & Analytics for AI Agents

What it is

Elasticsearch MCP is a Model Context Protocol server that connects AI agents to Elasticsearch clusters. It exposes search, indexing, and cluster management operations as MCP tools, letting agents like Claude Code run queries, manage indices, and analyze data through natural language instructions.

The server targets developers and data engineers who want their AI coding assistants to interact directly with Elasticsearch without writing raw REST calls.

How it saves time or tokens

Instead of manually constructing Elasticsearch DSL queries and copying results between terminal and editor, the MCP server lets your AI agent handle the full loop: build the query, execute it, and interpret results. This is especially useful for complex aggregations and index management tasks that require multiple API calls.

How to use

1. Install the Elasticsearch MCP server package.
2. Configure your Claude Code or other MCP client to connect to the server with your Elasticsearch cluster URL and credentials.
3. Ask your AI agent to query, index, or manage your Elasticsearch data in natural language.

Example

{
  "mcpServers": {
    "elasticsearch": {
      "command": "npx",
      "args": ["@elastic/mcp-server-elasticsearch"],
      "env": {
        "ES_URL": "https://localhost:9200",
        "ES_API_KEY": "your-api-key"
      }
    }
  }
}

Once configured, ask your agent: 'Search for error logs from the last hour in the app-logs index' and it constructs and executes the query.

Related on TokRepo

• MCP Integrations -- other MCP server integrations for AI agents
• AI Tools for Database -- database tools and search utilities

Common pitfalls

• The MCP server needs network access to your Elasticsearch cluster. Ensure firewall rules allow connections from your local machine.
• API key permissions scope what the agent can do. Use read-only keys for exploration and full-access keys only when you explicitly need write operations.
• Complex aggregation queries may exceed the agent's context window. Break large result sets into paginated requests.