abhigyanpatwari/GitNexus

GitNexus: The Zero-Server Code Intelligence Engine

Does it belong here? Yes

This is a genuine, substantial software project. It's not a list, tutorial, or demo — it builds a native code-indexing pipeline (Tree-sitter parsing, graph storage, MCP server, CLI, Web UI) designed to improve how AI coding agents understand code.

What it does (in plain terms)

AI coding assistants like Cursor, Claude Code, and Codex are good at writing code but often blind to the structure of a large codebase. They might edit one function without knowing 47 other functions depend on it, so "harmless" changes routinely break the build.

GitNexus fixes that by pre-building a knowledge graph of your entire codebase. Instead of letting the AI dig through raw files step-by-step, it indexes everything in advance: every file, function, dependency, call chain, and execution flow. It then exposes this graph to AI agents through MCP (Model Context Protocol) — a standard way to give AI agents extra tools.

How you use it

  • CLI + MCP (recommended): Run npx gitnexus analyze in your repo to index it. Run npx gitnexus setup to auto-connect your editor (Claude Code, Cursor, Codex, etc.). Running this writes configs, installs "skills" for the agent, and registers hooks that inject graph context automatically.
  • Web UI: A browser-based graph explorer and chat. It works entirely in-browser (via WASM) or in a "bridge mode" where it connects to your local CLI index without re-uploading.

The core idea: Precomputed Relational Intelligence

Traditional "Graph RAG" gives the AI raw graph edges and makes it query them one by one (e.g., Query 1: Find callers -> Query 2: What files? -> ... -> 4 queries later). GitNexus precomputes the answers. When you ask the AI "What depends on UserService?", the tool returns 8 callers, 3 clusters, and confidence scores in a single call. This makes answers reliable)Skip token-expensive multi-query chains) and makes smaller/cheaper LLMs work better because the heavy lifting is done by the indexer.

Key features (19 MCP tools + resources)

  • impact: Blast-radius analysis — lists everything that will break if you change a symbol.
  • trace: Shortest path between two functions.
  • detect_changes: Git-diff impact — maps changed lines to affected processes.
  • rename: Semantic rename (safe refactoring) with a dry-run preview.
  • explain / pdg_query: Program Dependence Graph (PDG) analysis for debugging statement-level flows and taint/source-sink paths.
  • Plus tools like query (hybrid semantic+keyword search), route_map (API endpoints), and read_file / grep.

It also generates agent skills (like Impact Analysis, Debugging, Refactoring) and installs hooks for Claude Code, Codex, and others so searches are auto-enriched with graph context before the agent acts.

Under the hood

  • Parsing: Tree-sitter (native for CLI, WASM for Web). Supports many languages; a few (Kotlin, Dart, Proto, Swift) use vendored grammars with prebuilt binaries.
  • Storage: LadybugDB (a fast native graph DB for CLI, WASM for browser).
  • Embeddings: Optional local semantic search via HuggingFace Transformers/ONNX (opt-in install).

Who is it for?

Engineering teams and individual developers using AI coding assistants on medium-to-large codebases who are tired of "AI hallucinated a bad edit" and want deeper, structural context for their tools.

Caveats & practical notes

  • License: PolyForm Noncommercial (it's free for non-commercial use, but you need a license for commercial use).
  • Setup quirks: Needs Node >= 22.15 for embedding features. npm 11 users may hit a crash; pnpm or global install is recommended.
  • Deployment: One-click deploy to Render runs about $35/month (server + web instance + disk).
  • Scale: CLI handles full repos of any size; the pure in-browser Web UI is limited to ~5k files (unless using the backend bridge).

Related

  • Project
    tirth8205/code-review-graph

    code‑review‑graph is a Python package that parses a repository with Tree‑sitter, stores the AST‑derived nodes/edges in a SQLite graph, and serves a minimal, blast‑radius slice of that graph to any MCP‑compatible AI coding assistant. It reduces per‑question token usage by ~63×, supports 30+ languages (including notebooks), offers incremental updates, framework‑aware edges for PHP/Laravel and Java/Spring, a GitHub Action for CI risk‑scored reviews, and optional semantic embeddings. Benchmarks show large token savings and an average F1 of 0.69 for impact analysis. The tool is aimed at developers and teams using LLM‑based code reviewers who need precise, low‑overhead context.

  • Project
    neo4j-labs/create-context-graph

    Create Context Graph is a CLI wizard that scaffolds a full‑stack AI application backed by a Neo4j knowledge graph. Choose a domain (e.g., healthcare), an LLM agent framework (Strands, PydanticAI, LangGraph, etc.), and a memory backend (hosted NAMS or self‑hosted Neo4j). The tool generates a FastAPI backend, a Next.js + Chakra UI frontend, Neo4j schema, optional demo data, and connectors for SaaS sources. It lets you prototype graph‑augmented agents quickly, with optional production‑grade self‑hosted mode for full Cypher and GDS support.

  • Project
    jeffhajewski/latticedb

    LatticeDB is an embedded, single‑file property‑graph database that also provides native HNSW vector similarity search and BM25 full‑text indexing. It offers a unified Cypher‑style query language, ACID transactions, and bindings for Python, TypeScript/Node, Go, and Java, making it suitable for local knowledge‑graph apps, agent memory, and RAG pipelines.

  • Project
    zereight/gitlab-mcp

    GitLab MCP Server is a lightweight, install‑once binary that exposes GitLab’s API as 275 MCP‑compatible tools (issues, merge‑requests, pipelines, wiki, releases, etc.) for AI agents. It supports PAT or OAuth authentication, multiple transports (stdio, SSE, streamable HTTP), fine‑grained permission modes, and can run locally, in Docker, or on custom GitLab instances. Designed for Claude, Codex, Cursor, VS Code, Copilot and other MCP clients, it lets LLM‑driven workflows manage GitLab resources without writing custom code.

  • Project
    ahmedkhaleel2004/gitdiagram

    GitDiagram is an AI-powered tool that automatically generates interactive architecture diagrams and narrated explainer videos for GitHub repositories, providing an MCP server for AI agents to query codebase structures.