About Code Context MCP Server
Code Context MCP Server is an MCP server that provides code context from local git repositories. It clones repositories, processes branches and files, generates embeddings for code chunks, and supports semantic search. It suits developers who want code retrieval over local repositories using Ollama embedding models.
Key features
- 01Clones local git repositories and processes branches and files
- 02Splits code into semantic chunks
- 03Generates code embeddings with Ollama
- 04Semantic code search across chunks
- 05Glob and keyword filtering of results
Use cases
- 01Finding relevant code in a repository by meaning
- 02Providing branch-specific code context to MCP clients
MCP support
This project is an MCP server that returns semantic search results from repositories through the queryRepo tool.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMECode Context MCP Server
A Model Context Protocol (MCP) server for providing code context from local git repositories. This server allows you to:
- Clone git repositories locally
- Process branches and files
- Generate embeddings for code chunks
- Perform semantic search over code
Features
- Uses local git repositories instead of GitHub API
- Stores data in SQLite database
- Splits code into semantic chunks
- Generates embeddings for code chunks using Ollama
- Provides semantic search over code
Prerequisites
- Node.js (v16+)
- Git
- Ollama with an embedding model
Installation
# Clone the repository
git clone <repository-url>
cd code-context-mcp
# Install dependencies
npm install
# Build the project
npm run build
Configuration
Set the following environment variables:
DATA_DIR: Directory for SQLite database (default: '~/.codeContextMcp/data')REPO_CACHE_DIR: Directory for cloned repositories (default: '~/.codeContextMcp/repos')
Using Ollama
For faster and more powerful embeddings, you can use Ollama:
# Install Ollama from https://ollama.ai/
# Pull an embedding model (unclemusclez/jina-embeddings-v2-base-code is recommended)
ollama pull unclemusclez/jina-embeddings-v2-base-code
Usage
Using with Claude Desktop
Add the following configuration to your Claude Desktop configuration file (claude_desktop_config.json):
{
"mcpServers": {
"code-context-mcp": {
"command": "/path/to/your/node",
"args": ["/path/to/code-context-mcp/dist/index.js"]
}
}
}
Tools
The server provides the following tool:
queryRepo
Clones a repository, processes code, and performs semantic search:
{
"repoUrl": "https://github.com/username/repo.git",
"branch": "main", // Optional - defaults to repository's default branch
"query": "Your search query",
"keywords": ["keyword1", "keyword2"], // Optional - filter results by keywords
"filePatterns": ["**/*.ts", "src/*.js"], // Optional - filter files by glob patterns
"excludePatterns": ["**/node_modules/**"], // Optional - exclude files by glob patterns
"limit": 10 // Optional - number of results to return, default: 10
}
The branch parameter is optional. If not provided, the tool will automatically use the repository's default branch.
The keywords parameter is optional. If provided, the results will be filtered to only include chunks that contain at least one of the specified keywords (case-insensitive matching).
The filePatterns and excludePatterns parameters are optional. They allow you to filter which files are processed and searched using glob patterns (e.g., **/*.ts for all TypeScript files).
Database Schema
The server uses SQLite with the following schema:
repository: Stores information about repositoriesbranch: Stores information about branchesfile: Stores information about filesbranch_file_association: Associates files with branchesfile_chunk: Stores code chunks and their embeddings
Debugging
MAC Mx Series - ARM Architecture Issues
When installing better-sqlite3 on Mac M-series chips (ARM architecture), if you encounter errors like "mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64e' or 'arm64')", you need to ensure the binary matches your architecture. Here's how to resolve this issue:
# Check your Node.js architecture
node -p "process.arch"
# If it shows 'arm64', but you're still having issues, try:
npm rebuild better-sqlite3 --build-from-source
# Or for a clean install:
npm uninstall better-sqlite3
export npm_config_arch=arm64
export npm_config_target_arch=arm64
npm install better-sqlite3 --build-from-source
If you're using Rosetta, make sure your entire environment is consistent. Your error shows x86_64 binaries being built but your system needs arm64. For persistent configuration, add to your .zshrc or .bashrc:
export npm_config_arch=arm64
export npm_config_target_arch=arm64
Testing Ollama Embeddings
curl http://localhost:11434/api/embed -d '{"model":"unclemusclez/jina-embeddings-v2-base-code","input":"Llamas are members of the camelid family"}' curl http://127.0.01:11434/api/embed -d '{"model":"unclemusclez/jina-embeddings-v2-base-code","input":"Llamas are members of the camelid family"}' curl http://[::1]:11434/api/embed -d '{"model":"unclemusclez/jina-embeddings-v2-base-code","input":"Llamas are members of the camelid family"}'
License
MIT
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.