MongoDB MCP

MongoDB MCP is the official Model Context Protocol server from MongoDB that connects AI-powered developer tools to MongoDB Atlas clusters, Community Edition, and Enterprise Advanced deployments. It lets AI agents explore databases, run queries, manage indexes, and perform CRUD operations through natural language, directly inside IDEs like VS Code, Cursor, and Windsurf. The server organizes its tools into three categories: Atlas tools for managing cloud resources (organizations, projects, clusters, database users), local Atlas tools for creating and managing local development clusters via the mongodb-atlas-local Docker image, and database tools for document operations, aggregation pipelines, and schema inspection. Recent updates added Performance Advisor integration so you can surface index recommendations and slow query diagnostics without leaving your editor. MongoDB MCP also supports vector search workflows. The insert-many tool can auto-generate embeddings using Voyage AI models for fields with vector search indexes, removing the manual embedding step. The CreateIndex tool handles both standard and vector search indexes through a single interface. For local development, the server can spin up ephemeral MongoDB clusters on demand, cutting setup time to seconds. The server is open source, runs via stdio or HTTP transport, and can be self-hosted or deployed in Docker.