The ultimate high-performance local code context engine for AI-native software development.
NakshAstraMCP is an ultra-fast, local-first Model Context Protocol (MCP) server built to empower AI coding assistants — including Claude, Cursor, Windsurf, and Antigravity IDE — with deep, AST-accurate, structural understanding of your codebases.
Unlike generic text searches or broad file dumps that inflate LLM token costs and dilute context quality, NakshAstraMCP parses class hierarchies, function boundaries, and cross-file reference graphs to supply AI agents with the exact context needed to solve complex programming tasks — safely and precisely.
| 🚀 Setup Guide | 📖 User Guide | 🤖 Agent Guide |
|---|---|---|
| Step-by-step Installation & Platform Setup | Advanced Usage, CLI Reference & Configuration | AI Agent Behavioral Instructions |
| 📜 License | 🛡️ Security Policy | 💬 Community Discussions |
|---|---|---|
| Usage Terms | Local Privacy & Safety | Q&A, Ideas & Feedback |
| 🎯 MCP-First Skill | 🛠️ Troubleshooting | 📋 Global Agent Config |
|---|---|---|
| AI Assistant Rule Profile | Diagnosis & Recovery | Custom AI Instructions Block |
- Search Latency: ~0.68ms (p95) across medium-to-large codebases.
- Semantic Alignment: CPU-bound FlashRank reranker aligns results to programming intent.
- Memory Footprint: Runs under < 150 MB idle RAM with automatic garbage collection.
- Language Coverage: Deep AST parsing for Python, JavaScript, TypeScript (TSX), Java, Kotlin, and Swift out of the box — plus runtime addon support for Go, Rust, Ruby, and more.
- Token Savings: Reduces LLM context payload sizes and costs by up to 75%.
Tested on a commercial codebase with over 10,000 source files:
| Metric | With NakshAstraMCP | Manual Search | Efficiency Gain |
|---|---|---|---|
| Context Fidelity | High: specific AST symbols & 1-hop neighbors | Low: scattered keyword-only search | High-Precision Context |
| LLM Token Cost | $0.09 | $0.37 | 75% Cost Reduction |
| Wall Clock Time | 1m 21s | 2m 05s | 35% Speed Increase |
Sleek multi-repository hybrid search with instant lexical routing.
| Feature | Description |
|---|---|
| 🔍 Multi-Repo Hybrid Search | Search and merge context across all your projects simultaneously. |
| 🧠 Semantic Reranking | FlashRank cross-encoder (CPU-based) re-orders results by conceptual intent. |
| 🌳 AST-Aware Snippets | Returns complete, syntactically valid functions/classes — no arbitrarily sliced text. |
| 📊 PageRank Relevance | Grades code importance based on cross-file call frequency and import patterns. |
| 🤖 Agent Orchestration | Auto-provisions AGENTS.md instructions and MCP-First Skill Profile for AI assistants. |
| 🛡️ Path Jail | Strictly sandboxed to registered workspace roots — protects sensitive environments. |
| 👁️ Real-Time Watcher | Debounced filesystem updates with automatic mass-update safeguards. |
| 🧩 Runtime Language Addons | Add new Tree-sitter grammars (e.g., Go, Rust) without rebuilding. |
| 🧹 Memory Guard | Background process prevents RAM leaks during long indexing runs. |
| 📈 Nebula UI Dashboard | Streamlit visualization tool for interactive codebase graph exploration. |
| 🌉 Dual Transport Bridge | Serve multiple AI clients simultaneously from a single background session. |
Real-time indexing statistics and memory usage tracking.
# Windows (PowerShell)
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"# macOS / Linux
curl -LsSf https://astral.sh/uv/install.sh | sh📥 Download v3.20.0 Secure Wheel from the Releases Page
# Recommended (uv) — auto-isolates the tool globally:
uv tool install https://github.com/vijaytank/NakshAstraMCP-Docs/releases/download/v3.20.0/nakshastramcp-3.20.0-cp313-cp313-win_amd64.whl --force# Navigate to your project root, then:
nakshastramcp start --workspace .nakshastramcp doctor # Runs 13 pre-flight environment diagnostics
nakshastramcp status # View active workspaces and server stateNext step: Connect NakshAstraMCP to your AI client. See the User Guide → Client Configuration.
| Tier | Hardware | Features Enabled |
|---|---|---|
| Minimal | 2 CPU Cores / 4 GB RAM | Core keyword search, aggressive Memory Guard cleanups |
| Recommended | 4 CPU Cores / 8 GB RAM | + Tantivy FTS, CPU FlashRank Semantic Reranking |
| Optimal | 8+ CPU Cores / 16 GB RAM | + PageRank graph calculations, deep AST relationship mapping |
NakshAstraMCP features a Dual Transport Bridge that lets multiple IDEs and AI clients share one background session with zero performance contention.
- Host Session: Your primary editor (e.g., Antigravity or Cursor) spawns the host via
stdiotransport. - HTTP Bridge: The host automatically exposes a streamable HTTP connection on port
2102. - Secondary Clients: Other tools (VS Code extensions, scripts) can connect simultaneously via:
- URL:
http://127.0.0.1:2102/mcp - Type:
streamable-http
- URL:
- 100% Local Execution: No code, index data, or metadata ever leaves your machine.
- Secret Detection: Integrated secret scanner blocks API keys and passwords from being indexed.
- Jailed Paths: Strict sandboxing blocks symbolic link exploits and out-of-workspace traversal.
- User-Space Only: Runs entirely within user-space — no administrator/root privileges required.
© 2026 Vijay Tank. All rights reserved.
🚀 Get Started · 📖 User Guide · 🛠️ Troubleshooting · 🛡️ Security · 💬 Discussions