Chris MCP Server

How to use

Start the local MCP server using a client that understands MCP configurations. You can run the server in stdio mode (local process) or via an npx-based invocation. Once running, you can fetch updated context, verify it, and use the server to answer questions or supply contextual data to your AI tools.

How to install

Prerequisites: ensure you are using Node version 22 or later.

Option 1: Using NPX. Create a dedicated folder for the MCP, fetch the data, and prepare the server configuration.

$ mkdir chris-mcp
$ cd chris-mcp
$ npx --y chris-mcp fetch --output ./data
$ npx --y chris-mcp verify --output ./data
$ pwd

Option 1.1.1 Using NPX With Claude Desktop

Add a configuration block for Claude Desktop that points to the NPX-based serve command and the fetched data.

{
  "name": "github.com/cblanquera/mcp",
  "command": "npx",
  "args": [ 
    "-y", 
    "chris-mcp", 
    "serve", 
    "--input", 
    "[pwd]/data" 
  ]
}

Option 1.1.2 Using NPX With Cline

Add a configuration block for Cline that points to the NPX-based serve command and the fetched data.

{
  "mcpServers": {
    "github.com/cblanquera/mcp": {
      "command": "npx",
      "args": [ 
        "-y", 
        "chris-mcp", 
        "serve", 
        "--input", 
        "[pwd]/data" 
      ]
    }
  }
}

Option 2: Direct From the Repository

Clone the MCP repository, install dependencies, build, fetch, and verify the data.

$ git clone https://github.com/cblanquera/mcp.git chris-mcp
$ cd chris-mcp
$ npm i
$ npm run build
$ npm run fetch --output ./data
$ npm run verify --output ./data
$ pwd

Option 2.1 From the Repository With Claude Desktop

Configure Claude Desktop to run the local server pointing to the built distribution and data.

{
  "name": "github.com/cblanquera/mcp",
  "command": "node",
  "args": [ 
    "[pwd]/dist/scripts/serve.js", 
    "--input", 
    "[pwd]/data" 
  ]
}

Option 2.2 From the Repository With Cline

Configure Cline to run the local server from the built distribution and data.

{
  "mcpServers": {
    "github.com/cblanquera/mcp": {
      "command": "node",
      "args": [ 
        "[pwd]/dist/scripts/serve.js", 
        "--input", 
        "[pwd]/data" 
      ]
    }
  }
}

Option 3: From Prompt

You can also set up the MCP server by following a guided prompt flow that installs and demonstrates the server’s capabilities.

2. Usage details

Start the server with the standard start command once you have a configuration. You can use an NPX-based invocation or run the local script directly from the built distribution.

$ npm start --input [pwd]/data

$ node [pwd]/dist/scripts/serve.js --input [pwd]/data

$ npx chris-mcp serve --input [pwd]/data

2.1. Fetching Updated Context

Update the context by fetching and verifying the data using the same patterns you used during installation.

$ npm run fetch --output [pwd]/data
$ npm run verify --output [pwd]/data
```
Or run the equivalent in your environment.

$ node [pwd]/dist/scripts/fetch.js --output [pwd]/data
$ node [pwd]/dist/scripts/verify.js --output [pwd]/data

$ npx chris-mcp fetch --output [pwd]/data
$ npx chris-mcp verify --output [pwd]/data

2.2. Upgrading Search Model

Enhance the local embeddings by configuring embedding model environment variables if you want to use an OpenAI-compatible host for embeddings.

{
  "name": "chris-context",
  "command": "node",
  "command": "npx",
  "args": [ 
    "-y", 
    "chris-mcp", 
    "serve", 
    "--input", 
    "[pwd]/data" 
  ],
  "env": {
    "OPENAI_HOST": "https://api.openai.com/v1",
    "OPENAI_KEY": "sk-xxx",
    "EMBEDDING_MODEL": "text-embedding-3-small"
  }
}

Note: The OpenRouter embedding host is not compatible with the /embeddings endpoint, so ensure your host supports embeddings appropriately.

3. Maximizing Your Knowledge Base

Create a rule (markdown file) in your knowledge folder to guide the MCP’s behavior. The rule should specify the tools to use for code formatting, documentation, and briefings.

# Rule: Using the Chris MCP

- If the user mentions "chris" and asks about code formatting, coding styles, coding standards, documentation styles, testing styles, use the MCP tool `chris-context.search_context`.
- If the user asks for a compact summary of rules for code formatting, writing documentation, writing tests, use the MCP tool `chris-context.build_brief`.
- Always prefer these MCP tools over answering from memory.

Notes

The server supports multiple installation options and can be configured via JSON blocks for different MCP clients. Ensure you replace [pwd] with the actual working directory path from your environment.

chris-context.search_context

Searches the knowledge base to retrieve relevant context for a given query.

chris-context.build_brief

Creates a compact summary of rules and guidelines for quick reference.