Qdrant RAG MCP Server

🆕 GitHub Integration (v0.3.0) + Projects V2 (v0.3.4) + Sub-Issues (v0.3.4.post4)

- 🤖 Intelligent Issue Resolution: RAG-powered GitHub issue analysis and automated fix generation
- 🔄 End-to-End Workflows: Analyze issues → Generate fixes → Create PRs with dry-run safety
- 📋 GitHub Projects V2 Management: Create projects, manage roadmaps, track progress with custom dashboards
- 🎯 29 GitHub MCP Tools: Complete issue lifecycle + 8 project management tools + 5 sub-issues tools via natural language
- 📑 Hierarchical Sub-Issues: Break down complex work with parent/child issue relationships

- List, add, remove, create, and reorder sub-issues
- Bulk add sub-issues to Projects V2 with smart field assignment
- Re-parenting support for reorganizing work
- Automatic issue number to ID conversion for intuitive usage

- 👁️ Context Window Visibility: Monitor what Claude knows in the current session
- 📊 Token Usage Tracking: Real-time estimates of context window consumption
- ⚠️ Usage Warnings: Automatic alerts at 60% and 80% context usage
- 📈 Session Timeline: Chronological view of all context-consuming operations
- 💾 Session Persistence: Automatic saving for later analysis
- 🔍 Session Viewer: Utility to analyze patterns across sessions

- 🎯 Content-Type Specific Models: Different AI models optimized for different content types

- Code: Programming-aware embeddings for 30-50% better code search relevance
- Config: Structure-aware models for precise JSON/YAML/XML navigation
- Documentation: Prose-optimized embeddings for natural documentation search

🆕 Progressive Context Management (v0.3.2)

- 🎯 Multi-Level Context: File → Class → Method hierarchy for optimal token usage
- 💰 50-70% Token Reduction: Get high-level overviews without sacrificing detail access
- 🧠 Query Intent Detection: Automatically determines the right context level for your query
- ⚡ Semantic Caching: Similar queries hit cache with 0.85 similarity threshold
- 🔄 Drill-Down Options: Start with overview, expand to details as needed
- 🏗️ Smart Structure Building: Automatically constructs code hierarchies from search results

- 🎯 Context-Aware: Automatically detects and scopes to your current project
- 🔍 Hybrid Search: Combines semantic understanding with keyword matching for +30% better precision
- 🧠 AST-Based Chunking: Structure-aware code parsing for Python, Shell, Go, JavaScript, and TypeScript (-40% tokens)
- 🔗 Dependency-Aware Search: Automatically includes files that import or are imported by your search results (v0.1.9)
- 📊 Enhanced Search Context: Get surrounding code chunks automatically for better understanding (v0.2.0)
- 🎯 Multi-Signal Ranking: 5-factor ranking system for 45% better search precision (v0.2.1)
- 📚 Documentation Indexing: Index and search markdown documentation files (v0.2.3)
- ⚡ Smart Incremental Reindexing: Only process changed files for 90%+ faster reindexing (v0.2.4)
- 📁 Multi-Project Support: Keep different projects' knowledge separate
- 🚀 Fast Local Execution: Supports Apple Silicon MPS acceleration
- 🔧 Specialized Indexers: Language-aware code parsing and config file understanding
- 🔄 Optional Auto-Indexing: Keep your index up-to-date automatically as files change
- 📊 Project-Aware Logging: Automatic log separation by project with rich debugging tools
- 🏥 Health Monitoring: Built-in health checks with detailed system status

Complete documentation for setting up and using the Qdrant RAG server with Claude Code.

- Claude Code CLI
- Docker
- Python 3.10+ withuv(ultraviolet package manager)

This project usesuvfor fast, reliable Python package management. If you don't have it installed:

# macOS/Linux curl -LsSf https://astral.sh/uv/install.sh | sh # Or with Homebrew brew install uv # Windows powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

# 1. Clone the repository git clone <repository-url> ~/mcp-servers/qdrant-rag cd ~/mcp-servers/qdrant-rag # 2. Run setup (starts Qdrant, installs dependencies) ./scripts/setup.sh # 3. Install globally with context awareness (ONE TIME ONLY) ./install_global.sh # That's it! The MCP server is now available in ALL projects # 4. Test in any project cd ~/any-project claude # Ask: "What's my current project context?"

# For current session only export QDRANT_RAG_AUTO_INDEX=true claude # For permanent auto-indexing echo 'export QDRANT_RAG_AUTO_INDEX=true' >> ~/.bashrc # or for zsh users: echo 'export QDRANT_RAG_AUTO_INDEX=true' >> ~/.zshrc

🔧 Working Directory Configuration (Important!)

The MCP server needs to know your actual working directory to correctly detect projects. You have three options:

Option 1: Natural Language (No Configuration Required!)

Simply tell Claude Code to set the working directory at the start of your session:

"Get current directory with pwd, export MCP_CLIENT_CWD to that value, then run health check"

This works immediately without any configuration changes!

Option 2: Environment Variable Configuration

In your Claude Code configuration (~/.claude-code/config.json):