MongoDB MCP Server

MongoDB databases queried and managed via schema-aware AI integration.

Data last scanned today · Reviewed today

688 tools★ 284TypeScriptMIT

Overview

This MCP server enables AI assistants to interact directly with MongoDB deployments. By exposing collection structures and automatically inferring schemas from sample documents, it allows for natural language database operations including standard queries, complex aggregations, and data modifications. It supports various infrastructure types, including standalone, replica sets, and Atlas clusters. The server features a configurable read-only mode to prevent accidental data changes, making it suitable for exploratory analysis. Additional developer utilities include automatic ObjectId string conversion, time-format normalization for accurate date queries, and support for both stdio and HTTP transport layers.

Our verdict

This server carries a high risk level with a security score of 48/100. It is community-maintained rather than official, and critical aspects such as the read-only implementation and authentication methods have not yet been editorially reviewed. Given these factors, users should exercise caution and avoid using this tool with production databases containing sensitive or mission-critical data until further verification is available.

Supports read-only mode to restrict write operations
Small supply-chain surface with only 4 dependencies
Provides automated schema inference from document samples
Supports complex aggregation pipelines and explain plans

Not maintained by MongoDB official vendor
Authentication mechanisms lack editorial security review
Read-only enforcement has not been independently verified

Setup

Claude Desktop

{
  "mcpServers": {
    "mcp-mongo-server": {
      "args": [
        "-y",
        "mcp-mongo-server"
      ],
      "command": "npx"
    }
  }
}

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop. Some servers require API keys or tokens as environment variables — check the project's README and add an "env" object if needed.

No read-only mode — running this grants full read/write access.
No built-in authentication — anyone with access to this config can invoke every tool it exposes.

Tools

Tool	Description	Risk
query	Execute MongoDB queries	low
aggregate	Run aggregation pipelines	low
count	Count matching documents	low
update	Modify documents	medium
insert	Add new documents	medium
createIndex	Create collection indexes	medium
serverInfo	Get MongoDB server details	low
convertTime	Convert a Unix timestamp or date string to UTC ISO 8601, GMT, and Unix seconds/milliseconds, and report the server's current time and timezone.	low

Compatibility

Client	Local	Docker	Remote	Read-only
Claude Desktop
Cursor
VS Code
Windsurf
ChatGPT

Frequently asked questions

›What are the software prerequisites for running this server?

This server requires Node.js version 20 or newer to run.

›How can I prevent the AI from accidentally modifying my data?

You can use the --read-only flag or the MCP_MONGODB_READONLY environment variable to block all write operations and risky aggregation stages.

›Does this server support remote access via HTTP?

Yes, you can use the --transport http flag to expose an endpoint, which allows for remote or multi-client access on a configurable port.

›How are security risks regarding cross-database access handled?

Aggregation stages like $out, $merge, or $lookup that target databases outside your connection string are blocked by default to prevent unauthorized data access.

›Are server-side JavaScript aggregation operators allowed by default?

No, operators like $function, $where, and $accumulator are rejected by default because they execute arbitrary code on the MongoDB server. You must explicitly enable them using the --allow-server-js flag.

›Can I secure the HTTP endpoint with an authentication token?

Yes, you can use the --auth-token flag or the MCP_HTTP_AUTH_TOKEN environment variable to require a Bearer token for all requests to the HTTP endpoint.