CJ-MCP Server

CJ-MCP is a minimal yet scalable MCP server that serves your project’s private knowledge and development templates to IDEs, enabling you to provide module-specific context and ready-to-use templates alongside language syntax references.

How to use

You run the CJ-MCP server locally and connect your MCP client (for example, Cursor) to it. The server exposes global resources like syntax and standard library references, plus module-specific documents under your project’s modules. When you work on a module, you can request targeted prompts or summaries to generate code, docs, or component templates that follow your project conventions. Context7 can be used alongside CJ-MCP to cover language syntax and standard libraries, while CJ-MCP handles project-specific knowledge and processes.

How to install

Prerequisites: Python 3.11 or newer is recommended. You will install dependencies and run the server locally.

# Option A: using conda
conda create -n mcp-env python=3.11 -y
conda activate mcp-env
pip install -r requirements.txt

# Option B: using a virtual environment
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# Run the MCP server from the project root
python server.py
# If successful, the server will wait for an MCP client to connect via stdio

Cursor configuration example

Create the MCP configuration in your project root at .cursor/mcp.json and add the following configuration. This example uses the local stdio approach with an explicit path to the server script.

{
  "mcpServers": {
    "cj-mcp": {
      "command": "python",
      "args": ["/path/to/my_mcp_server/server.py"],
      "env": {
        "PYTHONPATH": "/path/to/my_mcp_server"
      }
    }
  }
}

Notes on environment and paths

If you use a virtual environment, point command to the Python executable in that environment, for example: "/path/to/.venv/bin/python". Save your environment setup and ensure the PYTHONPATH points to the folder that contains server.py so CJ-MCP can import its modules.

What you get from CJ-MCP

Global resources such as cj://syntax and cj://global gather Markdown documents from the global directory. Module resources under proj://modules/{module} are auto-aggregated from your module’s documentation. You can request module-specific summaries, developer prompts, and component templates that adhere to your project’s conventions.

Troubleshooting tips

If the server does not appear connected in your MCP client, verify the Python path, ensure the server script path is correct, and restart Cursor after saving the .cursor/mcp.json file. Check that the virtual environment (if used) is active and that dependencies installed without errors.

Available tools

summarize_module

Summarizes a specified module's documentation into a concise, file-by-file digest to help you understand module scope and design.

cj_module_dev

Prompts the model to develop or modify a specific module by pulling global conventions and module docs, then generating code that follows project standards.

cj_module_summary

Generates a high-level summary of a module, including goals, key types, events, dependencies, and known pitfalls.

gen_cj_component

Generates a CJ component template based on a given name and context, aligning with project conventions.