Maximum Reasoning, Minimum Tokens. A privacy-first RAG engine that delivers surgical codebase context without wasting AI bandwidth.
What is RagCode?
RagCode enables AI Assistants (like Windsurf, Cursor, Antigravity, or Copilot) to instantly "understand" your entire project without reading thousands of lines of files. It uses Retrieval-Augmented Generation (RAG) mixed with deep Code Graph (AST) analysis to give the AI context exactly when it needs it.
Zero Cloud. Everything runs 100% locally on your machine using Qdrant and Ollama. Your proprietary code never leaves your laptop.
Dumping entire files into an AI's context window destroys its ability to think, whether you use Cloud models (Anthropic, OpenAI, Gemini) or Local models (Ollama, LM Studio, vLLM).
RagCode acts as a surgical filter: Instead of forcing the AI to read 15,000 lines of code, RagCode delivers only the precise 200-token AST chunks it needs.
- Local Models: Reclaim limited context windows for pure reasoning. Level the playing field and perform enterprise-grade codebase analysis with zero costs and 100% privacy.
- Cloud Models: Slash API costs by 95%, reduce input latency, and drastically minimize hallucinations.
Get started instantly in 5 minutes with our automated installer.
Linux (amd64) / WSL:
curl -fsSL https://github.com/doITmagic/rag-code-mcp/releases/latest/download/rag-code-mcp_linux_amd64.tar.gz | tar xz && ./rag-code-install -ollama=docker -qdrant=dockermacOS (Apple Silicon):
curl -fsSL https://github.com/doITmagic/rag-code-mcp/releases/latest/download/rag-code-mcp_darwin_arm64.tar.gz | tar xz && ./rag-code-install -ollama=docker -qdrant=dockerWindows (PowerShell):
Invoke-WebRequest -Uri "https://github.com/doITmagic/rag-code-mcp/releases/latest/download/rag-code-mcp_windows_amd64.zip" -OutFile "ragcode.zip"
Expand-Archive ragcode.zip -DestinationPath . -Force
.\rag-code-install.exe -ollama=docker -qdrant=dockerRead the Full QUICKSTART Guide for your first AI Prompt
RagCode has evolved into a massively powerful engine. Choose your path:
- The Vibe Coder Path (Quickstart) - I just want it to work in my IDE right now.
- The Developer Path (Contributing) - I want to build RagCode locally or contribute code.
- The AI Agent Path (Headless/HTTP) - I am an AI or script trying to query the workspace directly without an IDE.
- The Architect Path (Docs) - Daemon/Adapter IPC, Workspace Isolation, AST Parsing, Hybrid 60/40 Scoring, and Lazy Stale Cleanup.
RagCode V2 isn't just a vector database wrapper. It features deep language understanding:
- Zero-Wait Fallback AST Search: If your codebase is still indexing, RagCode falls back to lexical/AST structural search so you never wait to get work done.
- Path-Scoped Boosts: The engine automatically detects what file your AI is currently editing and boosts search results from the same folder or related logic.
- Nested Workspace Detection: Monorepos and deeply nested git submodules are handled safely through a unified detection registry.
- Telemetry & JSONL Insights: RagCode analyzes exactly how much context your AI saves and records precise match reasons for why a snippet was chosen.
- Skill Ecosystem (
rag_list_skills): Enhance your codebase on-the-fly. Agents can install custom behavioral skills into.ragcode/skillsto expand their own capabilities dynamically.
- Go: Complete native AST support
- PHP: Vanilla PHP, Laravel, WordPress (Hooks, Widgets, WooCommerce, Oxygen Builder)
- JavaScript & TypeScript: Vanilla JS/TS, Node.js, React, React Native, Next.js, Vue
- Python: Complete native AST support
- HTML & Markdown: Structural documentation mappings
- Generic Support: CSS, JSON, YAML, Shell scripts, SQL
Built for developers who want smarter AI code assistants
Star us on GitHub if RagCode speeds up your workflow!