pycodesage 0.2.0

Local-first CLI code intelligence tool with LangChain-powered RAG

CodeSage

Local-first code intelligence CLI with MCP support for Claude Desktop, Cursor, and Windsurf.

Index your codebase and search it using natural language. Everything runs locally with Ollama. Supports Python, JavaScript, TypeScript, Go, and Rust.

Install

pip install pycodesage

Or from source:

git clone https://github.com/keshavashiya/codesage.git
cd codesage
pip install -e .

Requirements

Ollama must be running with these models:

ollama pull qwen2.5-coder:7b
ollama pull mxbai-embed-large
ollama serve

Usage

# Initialize and index your project
cd your-project
codesage init
codesage index

# Search your code
codesage suggest "validate email"

# Check everything is working
codesage health

MCP Setup

{
  "mcpServers": {
    "codesage": {
      "command": "codesage",
      "args": ["mcp", "serve", "--global"]
    }
  }
}

Other MCP clients (Cursor, Windsurf)

Cursor: Settings → Features → MCP Servers, add same config.

Windsurf: Settings → MCP → Add Server. Command: codesage, Args: mcp serve --global

Commands

codesage init           # Initialize project
codesage index          # Index codebase
codesage suggest QUERY  # Search code
codesage stats          # Show stats
codesage health         # System check
codesage review         # AI code review
codesage chat           # Interactive mode

More commands

# MCP
codesage mcp serve          # Start server
codesage mcp serve --global # All projects
codesage mcp test           # Test tools

# Security
codesage security scan      # Scan vulnerabilities
codesage hooks install      # Pre-commit hook

# Storage
codesage storage info       # Backend details
codesage storage stats      # Metrics

# Profile
codesage profile show       # Developer profile
codesage profile patterns   # Learned patterns

Configuration

Stored in .codesage/config.yaml:

project_name: my-project
languages:
  - python
  - typescript

llm:
  provider: ollama
  model: qwen2.5-coder:7b
  embedding_model: mxbai-embed-large

exclude_dirs:
  - venv
  - node_modules
  - .git

Development

pip install -e ".[dev]"
pytest tests/ -v

License

MIT