Nx MCP
Nx MCP server: the agent sees monorepo projects, reads Nx docs and inspects failed CI tasks in Nx Cloud
- 1,415 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The server reads the workspace configuration and project graph, and through Nx Cloud it gets CI task output
- `update_self_healing_fix` applies or rejects a CI fix, so such a call is better confirmed by hand
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/nx-mcpDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The server reads the workspace configuration and project graph, and through Nx Cloud it gets CI task output
- `update_self_healing_fix` applies or rejects a CI fix, so such a call is better confirmed by hand
This is third-party code. Review the repository files before installing.
What it does
Nx MCP is a server from the Nx Console repository that hands an agent facts about an Nx monorepo: project relationships, tasks, code owners, generators and documentation. With that context the agent writes code for your stack and judges which projects a change touches. The default minimal mode exposes `nx_docs`, `nx_current_running_tasks_details`, `nx_current_running_task_output` and `nx_visualize_graph`. The workspace analysis tools (`nx_workspace`, `nx_project_details`, `nx_generators` and three more) come back with the `--no-minimal` flag. When a workspace uses Nx Cloud, `ci_information`, `ci_task_output` and `update_self_healing_fix` appear. The agent reads failed CI tasks with their output and applies, rejects or reruns the fix Nx Cloud proposed. Recent CI runs are also served as MCP resources.
Who it is for. Developers of Nx monorepos who work with an agent in Claude Code, Cursor or VS Code and want it to see the workspace structure.
Good fit when
- An agent edits a shared library and has to find which apps depend on it
- For a new project the agent picks an Nx generator and reads its parameter schema (with the `--no-minimal` flag)
- A build fails in CI on Nx Cloud and the agent reads the failed task output itself
- A question about Nx features gets an answer from the documentation sections `nx_docs` returns
Not a fit when
- A workspace without Nx: answers rest on the Nx project graph, so the server has nothing to work with
- CI data is wanted without Nx Cloud: `ci_information` and `ci_task_output` do not appear in such a workspace
- The agent needs the full analysis set at once: without `--no-minimal` part of the tools stays hidden
Example request
Which monorepo projects does a change in libs/shared/ui touch, and which tasks have to rerun?MCP
- Transport
- stdio, http
- Authentication
- not required
Limitations
- CI tools work only with Nx Cloud enabled.
- The default minimal mode hides six workspace analysis tools; `--no-minimal` brings them back.
- `nx_visualize_graph` needs a running IDE.
- Without a workspace path only `nx_docs` and `nx_available_plugins` are available.
- The SSE transport is removed; over a network the server speaks only Streamable HTTP, on port 9921 by default.
- The Nx Console extension manages the server itself only in Cursor; other clients add it by hand.
How to disable. `claude mcp remove nx-mcp` removes the server from Claude Code. In other clients delete the `nx-mcp` entry from `mcp.json`. Server telemetry is switched off with `"disableTelemetry": true` in `.nx/nx-mcp-config.json`.
FAQ
How do I connect Nx MCP to Claude Code?
Run `claude mcp add nx-mcp npx nx-mcp@latest`. The server starts through `npx` over stdio and needs no keys. Open Claude Code in the Nx workspace directory and ask it to describe the projects.
Why does the agent not see nx_workspace and nx_generators?
Those tools are hidden in minimal mode, which is on by default. Start the server with the flag: `npx nx-mcp@latest --no-minimal`. All six analysis tools then show up in the list.
How do I run Nx MCP over HTTP?
Add `--transport http`; `--port` sets the port, 9921 by default. A client connects to `http://localhost:<port>/mcp` over Streamable HTTP, and several clients can work at once. The SSE transport is gone.
Does it need Nx Cloud?
No, documentation, running tasks and the graph work without it. `ci_information`, `ci_task_output`, `update_self_healing_fix` and the CI run resources appear only in a workspace with Nx Cloud.
How do I configure the server without command-line arguments?
Put `.nx/nx-mcp-config.json` in the workspace root. The keys `tools`, `minimal`, `debugLogs`, `disableTelemetry`, `transport` and `port` mirror the CLI options, and command-line arguments win over the file.
Discussion
No questions or reviews yet.
Related
MCP serversA transparent proxy between an AI client and an MCP server: shows every JSON-RPC frame in the terminal, checks a capture in CI and exports spans
An MCP server that lets an agent read files, search code and run commands in a workspace
An MCP server that builds a code graph once and serves the answer to the agent instead of re-reading files
An MCP server in one binary: AI reads and edits files across several projects and runs your commands
add-mcp writes an MCP server into the configs of 25 agents with one command: Claude Code, Codex, Cursor, VS Code, Zed and more
A command-line debugger for MCP servers from Giant Swarm: a REPL for tools, OAuth 2.1 and an MCP server mode for Cursor
README in short
The README opens with what the server does and two ways to use it: the `nx-mcp` package through `npx`, or the Nx Console extension, which manages the server by itself in Cursor. It then covers minimal mode, the tool list, Nx Cloud resources and the config file. The contributor section explains building with `nx run nx-mcp:build` and testing in MCP Inspector.