A powerful, fully local AI coding assistant tailored for software engineers. Run high-performance models (like qwen2.5-coder) directly on your hardware (optimized for NVIDIA 3060+) without data ever leaving your machine.

No API costs. No internet required. Full privacy.

• 🏎️ Optimized Performance: Leverages Ollama for fast, local inference.
• 🛠️ Tool-Augmented: The agent can search your repository, run tests, and even access web documentation.
• 🔌 Seamless Integration: Works out-of-the-box with VS Code, Windsurf, and the Claude Code CLI.
• 🔍 RAG Powered: (Phase 5) Smart context retrieval from your codebase and web search results.
• 🛡️ Privacy First: Your code never leaves your local network.

graph TD
    VSCode[VS Code Extension] -->|port 4000| Agent[Agent Backend]
    CLI[Claude Code CLI] -->|port 3000| Proxy[Anthropic Proxy]
    Windsurf[Windsurf] -->|port 3000| Proxy
    
    subgraph "Internal Agent Tools"
        Agent --> Search[searchRepo]
        Agent --> Test[runTests]
        Agent --> Web[webSearch/RAG]
    end
    
    Agent -->|Claude API| Proxy
    Proxy -->|Ollama API| Ollama[Ollama Engine]
    Ollama --> Model[qwen2.5-coder:7b]

Ensure you have Ollama installed and running.

# Pull the recommended model
ollama pull qwen2.5-coder:7b

Run the automated startup script to launch the proxy and agent backend.

cd "c:\My Project\Claude_local_setup"
.\scripts\start-all.ps1

Run the health check script to ensure the proxy is responding correctly.

.\scripts\test-proxy.ps1

1. Open the vscode-extension folder in a new VS Code window.
2. Run npm install && npm run compile.
3. Press F5 to start the extension development host.
4. Right-click any code snippet to use Explain, Refactor, or Implement.

Point the CLI to your local proxy:

$env:ANTHROPIC_BASE_URL = "http://localhost:3000"
$env:ANTHROPIC_API_KEY  = "local-model"
claude

See project_structure.md for a deep dive into the architecture.

AGENT_PORT=4000
PROXY_BASE_URL=http://localhost:3000
WORKSPACE_ROOT=C:\My Project\Claude_local_setup
BRAVE_API_KEY= # Optional for web search

OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=qwen2.5-coder:7b
PROXY_PORT=3000

To safely shut down all backend processes, run:

.\scripts\stop-all.ps1

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the Project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Commit your Changes (git commit -m 'Add some AmazingFeature')
4. Push to the Branch (git checkout origin feature/AmazingFeature)
5. Open a Pull Request

Built with ❤️ for the developer community.