NakshAstraMCP-Docs

The ultimate high-performance code context engine for AI-native development.

📖 Overview

NakshAstraMCP provides AI agents (Claude, Cursor, Antigravity, etc.) with deep, structural understanding of your local codebase. Using advanced AST parsing and high-speed semantic ranking, it delivers the exact context your developer tools need to solve complex problems across large projects.

🗺 Documentation Hub

Quickly navigate to detailed guides:

🏆 Assessment & Performance

• Overall Efficiency: 9.3 / 10 (Industry-leading structural context)
• Index/Search Latency: ~0.68ms p95 (Ultra-low latency on 10,000-file repos)
• Idle Memory: < 150 MB RAM
• Language Support: Deep AST integration for Python, JavaScript, TypeScript, Java, and Kotlin.
• ROI Benchmark: 75% Cost Reduction vs. manual LLM context retrieval.

📊 Performance vs. Manual Analysis

Lightning-fast hybrid search across multiple repositories.

✨ Key Features

• 🔍 Hybrid Multi-Repo Search — Indexed search across all your projects simultaneously.
• 🧠 Semantic Reranking — AI-powered results prioritized by conceptual relevance using FlashRank.
• 🌳 AST Call Graph & Lexical Call Resolution — Natively extracts classes and functions, building a semantic call-graph using scope-constrained lexical boundary checks.
• 🤖 Automated Agent Orchestration — Self-provisioning AGENTS.md instructions with non-destructive backup.
• 🗺️ High-Fidelity Knowledge Mapping — Automated reports featuring distinct high-impact file and symbol matrices with clickable, precise line-highlight links (e.g. file:///#Lstart-Lend) for instant IDE navigation.
• 📦 Clean Louvain Clusters — Groups tightly-coupled files into functional modules named by dominant parent directories, excluding external import stubs for pure context.
• 🩺 Surgical Intelligence — High-precision tools (read_file, find_symbol, find_references) for localized context retrieval.
• 🛡️ Administrative Control — Full CLI control over server lifecycle (stop, restart, logs).
• 👁️ Real-Time Watcher — Changes are indexed instantly with mass-update protection.
• 🧩 Runtime Language Addons — Provision new Tree-sitter grammars at runtime.
• 📈 Visual Dashboard — Interactive Nebula Graph UI with health monitoring.
• 🧹 Operational Resilience — Built-in Memory Guard and WAL checkpointing for stability.

Detailed indexing statistics and server health monitoring.

🚀 Quick Start (Fast-Track)

1. Unified Installation

Requires uv. Install the secure binary wheel directly:

📥 Download v3.13.0 Secure Wheel (Windows)

uv tool install https://github.com/vijaytank/NakshAstraMCP-Docs/releases/download/3.0.0/nakshastramcp-3.13.0-cp313-cp313-win_amd64.whl --force

or if you get any errors try

python -m pip install .\nakshastramcp-3.13.0-cp313-cp313-win_amd64.whl

2. Register & Index

Initialize your workspace roots to build the local knowledge graph:

nakshastramcp start --workspace C:\path\to\your\project

3. Verification & Health

nakshastramcp status  # Check indexing progress
nakshastramcp doctor  # Perform full environment audit
nakshastramcp report . # Generate architectural knowledge map

💻 Hardware Tiers

NakshAstraMCP adapts to your system automatically:

🌉 Multi-Client Connectivity

NakshAstraMCP supports concurrent sessions from multiple IDEs via the Dual Transport Bridge.

• Primary Host: Your main IDE (e.g., Cursor) starts the host session.
• HTTP Follower: Configure secondary tools (e.g., VS Code extension) to connect to the bridge:
  • URL: http://127.0.0.1:2102/mcp
  • Type: streamable-http

🛡 Security & Privacy

• 100% Local: No source code or indices ever leave your machine.
• Sensitive Data Detection: Integrated secret scanner prevents indexing of API keys.
• Sandboxed Execution: The engine only accesses registered workspace roots.
• Zero Telemetry: No usage data is collected. Fully offline operation.