# MongoDB > Query and inspect MongoDB collections from Claude, Cursor, and VS Code. [Canonical HTML page](https://top-mcps.com/mcp/mongodb) · [server.json](https://top-mcps.com/mcp/mongodb.json) · [methodology](https://top-mcps.com/about/methodology) ## Install ### Claude Desktop — `claude_desktop_config.json` Paste under mcpServers. Fully quit and reopen Claude after editing. ```json { "mcpServers": { "mongodb": { "command": "npx", "args": [ "-y", "mongodb-mcp-server", "--connectionString=\"mongodb+srv://user:pass@cluster.mongodb.net\"" ], "env": { "MDB_CONNECTION_STRING": "${MDB_CONNECTION_STRING}" } } } } ``` ### Claude Code — `CLI or .mcp.json` Run from your repo. Commit .mcp.json to share with your team. ```shell # export MDB_CONNECTION_STRING=mongodb+srv://user:pass@cluster.mongodb.net claude mcp add mongodb -- npx -y mongodb-mcp-server --connectionString="mongodb+srv://user:pass@cluster.mongodb.net" ``` ### Cursor — `.cursor/mcp.json` Global path: ~/.cursor/mcp.json. Reload window after editing. ```json { "mcpServers": { "mongodb": { "command": "npx", "args": [ "-y", "mongodb-mcp-server", "--connectionString=\"mongodb+srv://user:pass@cluster.mongodb.net\"" ], "env": { "MDB_CONNECTION_STRING": "${MDB_CONNECTION_STRING}" } } } } ``` ### VS Code — `.vscode/mcp.json` VS Code uses the "servers" key (not "mcpServers"). ```jsonc { "servers": { "mongodb": { "command": "npx", "args": [ "-y", "mongodb-mcp-server", "--connectionString=\"mongodb+srv://user:pass@cluster.mongodb.net\"" ], "env": { "MDB_CONNECTION_STRING": "${MDB_CONNECTION_STRING}" } } } } ``` ### Windsurf — `~/.codeium/windsurf/mcp_config.json` Open via Cascade → hammer icon → Configure. ```json { "mcpServers": { "mongodb": { "command": "npx", "args": [ "-y", "mongodb-mcp-server", "--connectionString=\"mongodb+srv://user:pass@cluster.mongodb.net\"" ], "env": { "MDB_CONNECTION_STRING": "${MDB_CONNECTION_STRING}" } } } } ``` ### Cline — `cline_mcp_settings.json` Open via the Cline sidebar → MCP Servers → Edit. ```json { "mcpServers": { "mongodb": { "command": "npx", "args": [ "-y", "mongodb-mcp-server", "--connectionString=\"mongodb+srv://user:pass@cluster.mongodb.net\"" ], "env": { "MDB_CONNECTION_STRING": "${MDB_CONNECTION_STRING}" } } } } ``` ### Continue — `~/.continue/config.json` Continue uses modelContextProtocolServers with a transport block. ```json { "experimental": { "modelContextProtocolServers": [ { "transport": { "type": "stdio", "command": "npx", "args": [ "-y", "mongodb-mcp-server", "--connectionString=\"mongodb+srv://user:pass@cluster.mongodb.net\"" ], "env": { "MDB_CONNECTION_STRING": "${MDB_CONNECTION_STRING}" } } } ] } } ``` ### Codex CLI — `~/.codex/config.toml` Codex uses TOML. Each server is a [mcp_servers.] subtable. ```shell # ~/.codex/config.toml [mcp_servers.mongodb] command = "npx" args = [ "-y", "mongodb-mcp-server", "--connectionString=\"mongodb+srv://user:pass@cluster.mongodb.net\"", ] env = { MDB_CONNECTION_STRING = "${MDB_CONNECTION_STRING}" } ``` ### Zed — `~/.config/zed/settings.json` Zed calls them "context_servers". Settings live-reload on save. ```jsonc { "context_servers": { "mongodb": { "command": { "path": "npx", "args": [ "-y", "mongodb-mcp-server", "--connectionString=\"mongodb+srv://user:pass@cluster.mongodb.net\"" ] }, "env": { "MDB_CONNECTION_STRING": "${MDB_CONNECTION_STRING}" } } } } ``` ### ChatGPT — `ChatGPT → Apps directory` MongoDB doesn't ship a hosted HTTPS endpoint today. ChatGPT supports remote MCP servers only — to use this server in ChatGPT you'll need to deploy it to a public HTTPS URL first (e.g. via Cloudflare Workers or Vercel) or wait for an official remote build. ```none ``` ## At a glance - **Maintainer:** MongoDB - **Transport:** stdio - **Auth model:** API key - **Required secrets:** MDB_CONNECTION_STRING - **Supported clients:** Claude, Cursor, VS Code, Windsurf, Any MCP-compatible client, MongoDB 5+, MongoDB Atlas - **License:** Apache-2.0 - **Language:** TypeScript - **Latest version:** latest - **Last verified:** 2026-05-27 - **GitHub stars:** 1,038 (fetched 2026-06-02T11:55:52.290Z) - **Score:** 89/100 (rubric 2026-04 — see https://top-mcps.com/about/methodology) - **Source:** https://github.com/mongodb-js/mongodb-mcp-server ## Tools & permissions | Tool | Description | Args | Side effects | |------|-------------|------|--------------| | `list_databases` | List databases on the cluster. | `—` | read | | `list_collections` | List collections within a database. | `database: string` | read | | `find` | Run a find query against a collection. | `collection: string, filter?: object` | read | | `aggregate` | Run an aggregation pipeline. | `collection: string, pipeline: array` | read | ## Security & scope - **Access scope:** read-only - **Sandbox:** Connects via standard MongoDB driver to the connection string in MDB_CONNECTION_STRING. Read-only mode disables write tools at the MCP layer; combine with a database user scoped to the smallest namespace your workflow needs. - **Gotchas:** - Read-only is enforced at the tool level, not the MongoDB role. Always pair MCP read-only mode with a read-only database user. - Schema inference samples N documents; rare fields may be missing from the inferred schema. - The Atlas Admin tools require a separate Atlas API key — do not reuse the connection-string password for admin scope. ## Quick answer **What it does.** Connects to a MongoDB deployment and exposes database/collection enumeration, schema inspection, find queries, and aggregation pipelines to AI models. **Best for:** - Collection schema exploration - Aggregation pipeline drafting - Document inspection - Reporting queries - Index review **Not for:** - Transactional multi-document writes - Long-running Change Streams - Sharded-cluster ops work ## Description The official MongoDB MCP connects an AI agent to a MongoDB cluster — list databases, inspect collection schemas, run aggregations, and execute find/count queries. Supports Atlas, self-hosted MongoDB, and connection-string-based auth. The server ships read-only and write-capable modes; the safe default is read-only for any agent-driven conversation. ## Why it matters MongoDB is the most-installed document store. Wiring it into an agent unblocks schema exploration, ad-hoc reporting, and the "what is in this collection again" workflow without breaking out a separate client. ## Key features - Atlas + self-hosted support - Connection-string auth - Schema inference per collection - Read-only mode by default - Aggregation pipeline execution - Index introspection ## FAQ ### Does it support MongoDB Atlas? Yes — pass the standard SRV connection string from the Atlas UI. The MCP also wraps Atlas Admin endpoints when you provide an Atlas API key, but you can run pure-data workflows without those scopes. ### Read-only or read-write? Both. Read-only is the safe default. Pass --readOnly=false (or set MDB_READ_ONLY=false) to enable insert/update/delete tools — only do this on dev clusters. ### How does it infer schema? It samples a configurable number of documents per collection and returns the union of field types. Useful for the model, but treat the result as a sketch — not a contract. ### Can I scope to one database? Yes. Pass --database= to restrict every tool to that database. Combine with a database user that only has access to that namespace as defence in depth. ### Does it support transactions? No agent-facing transaction tool by design — multi-statement consistency in an LLM conversation is a footgun. Use a regular MongoDB driver for transactional workloads. ## Changelog - **2026-05-27** — Refreshed install snippets and fact sheet; verified for 2026. - **2025-04-22** — Initial directory listing.