MongoDB

• 3.2.1 Latest Nov 11, 2026

  mcp-mongo-server 3.2.1 fixes a query timeout regression in fresh npm installations of 3.2.0.
  
  The package now pins the MongoDB driver to 7.5.0 so configured query timeouts are enforced when MongoDB is slow to respond. Regression tests cover stdio and Streamable HTTP against both a standalone server and a replica set.
  
  Update the package version in your MCP client configuration to `[email protected]`.
  
  [Full changelog](https://github.com/kiliczsh/mcp-mongo-server/compare/3.2.0...3.2.1)
  

  View on GitHub ↗
• 3.2.0 Nov 11, 2026

  mcp-mongo-server 3.2.0 adds collection inspection tools and configurable query limits.
  
  - Use `listIndexes` to inspect index keys and options, and `collectionSchema` to infer fields from sample documents. Both accept the existing connection and database selectors.
  - Set optional document, UTF-8 result byte, and operation time limits for queries and aggregations. Set `query.includeTotal` to `false` to skip the exact count while keeping pagination metadata.
  - Write tools appear in discovery when at least one configured connection is writable. Use `listConnections` to choose a writable target.
  
  Update the package version in your MCP client configuration to `[email protected]`. See the [tool guide](https://github.com/kiliczsh/mcp-mongo-server/blob/3.2.0/docs/tools.md) and [configuration options](https://github.com/kiliczsh/mcp-mongo-server/blob/3.2.0/README.md#environment-variables).
  
  [Full changelog](https://github.com/kiliczsh/mcp-mongo-server/compare/3.1.0...3.2.0)
  

  View on GitHub ↗
• 3.1.0 Nov 11, 2026

  mcp-mongo-server 3.1.0 adds named MongoDB connections. One MCP server can now query several MongoDB deployments and select a database for each request.
  
  - Configure connections with `MCP_MONGODB_CONNECTIONS` and discover available targets with `listConnections` and `listDatabases`.
  - Pass `connectionId` and `database` to choose a target. Each connection can limit its allowed databases and run in read-only mode.
  - Existing single-URI configurations continue to work. You can keep `MCP_MONGODB_URI` as the default and add named connections alongside it.
  
  Update the package version in your MCP client configuration to `[email protected]`. See the [named connection example](https://github.com/kiliczsh/mcp-mongo-server/blob/3.1.0/README.md#named-connections) for setup.
  
  [Full changelog](https://github.com/kiliczsh/mcp-mongo-server/compare/3.0.2...3.1.0)
  

  View on GitHub ↗
• 3.0.2 Nov 11, 2026

  ## Changes
  
  - Block server-side JavaScript in query, count, and update filters by default. Read-only mode keeps the block active even when server-side JavaScript is explicitly enabled.
  - Clarify MCP tool descriptions so clients can choose the right MongoDB operation.
  - Fix collection filters, support string and numeric index commit quorum values, and correctly encode collection resource URIs.
  - Add a Docker test environment with standalone and single-node replica set fixtures.
  
  Full changelog: https://github.com/kiliczsh/mcp-mongo-server/compare/3.0.1...3.0.2
  

  View on GitHub ↗
• 3.0.1 Jul 29, 2026

  **Patch release.**
  
  Renames the CLI bin from `mongodb` to `mcp-mongo-server` so `npx -y mcp-mongo-server "<connection-string>"` (the form documented in the README) resolves directly.
  
  Previously, because the package name (`mcp-mongo-server`) and bin name (`mongodb`) differed, `npx` failed to resolve the single bin:
  
  ```
  sh: mongodb: command not found
  ```
  
  and the only working invocation was the awkward `npx -y -p mcp-mongo-server mongodb "<connection-string>"`.
  
  No API or behavior changes.
  
  **Full changelog:** https://github.com/kiliczsh/mcp-mongo-server/pull/45
  

  View on GitHub ↗

No comments yet. Sign in to start the discussion.

More Data servers

• Filesystem

  Read, write, and search files on the local filesystem with configurable access boundaries.

• Postgres

  Read-only access to a Postgres database — list tables, run SELECTs, inspect schemas.

• SQLite

  Read + write access to a local SQLite database file — full CRUD, schema management.