- Category
- Developer Tools
- Rank
- No. 98Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- abhigyanpatwari
- GitHub
- 47.1k stars
- Latest release
- v1.6.12-rc.8
- Date
About
A zero-server code intelligence engine that creates interactive knowledge graphs from any codebase. Runs entirely client-side in your browser or via CLI, mapping every dependency, call chain, and execution flow to give AI agents deep architectural context.
What it does
GitNexus analyzes source code ahead of time, records relationships between symbols and files, then answers focused architecture questions. Developers can inspect results through a visual browser interface or connect coding agents through MCP. Its queries cover search, symbol context, dependency impact, paths between symbols, and changed-code effects.
Why it's ranked here
The project addresses a real weakness in coding agents: incomplete awareness of downstream dependencies. Precomputing clusters, traces, and confidence before queries can reduce repeated exploration. Local storage and broad editor integration strengthen the case. The verdict is tempered by native installation complexity, browser memory limits, and a noncommercial license that restricts some adoption.
What's good
The CLI keeps repositories and persistent indexes local, with no network required. The browser option stores its in-session graph locally and needs no installation. Precomputed structure gives agents categorized references and blast-radius information without forcing them to assemble many raw edges. Bridge mode also lets the visual interface reuse CLI indexes instead of uploading and indexing repositories again.
Tradeoffs
Browser analysis is limited by memory to roughly 5,000 files unless a backend is used. Large hosted indexes are memory-bound. Native database and parser dependencies can complicate installation, especially with cold package caches or npm 11. Skipping native grammar builds removes parsing for Dart, Proto, Swift, and Kotlin. Optional embeddings have additional Node version and runtime-install constraints. The PolyForm Noncommercial license is also a material restriction.
How to use it well
Use the CLI and MCP integration for daily work on repositories where agents must understand callers, dependencies, and change impact before editing. Use the browser interface for quick exploration, demonstrations, and one-off questions. Prefer bridge or backend mode for repositories beyond browser memory limits. Treat GitNexus as context infrastructure for an editor or coding agent, not as a substitute for tests, review, or production access controls.
Technical notes+
The root package.json defines a private monorepo with ESLint, Prettier, Husky, and GitNexus refresh scripts. gitnexus/vitest.config.ts splits native LadybugDB, CLI end-to-end, and default tests, serializing native database suites and enforcing coverage floors. gitnexus-web/vitest.config.ts uses jsdom, while gitnexus-web/playwright.config.ts runs Chromium with software WebGL. eslint.config.mjs protects MCP stdout framing, database shutdown behavior, and safe Tree-sitter parsing through eslint-rules/require-safe-parse.mjs. docker-server.mjs serves static assets and proxies API requests, requiring a minimum 32-character bearer token for public proxy mode and comparing credentials in constant time. docker-server.test.mjs exercises traversal rejection, configuration injection, script escaping, authentication, and proxy behavior.
Observed
- License
- PolyForm Noncommercial 1.0.0
- Primary language
- TypeScript, with JavaScript server and Python evaluation tooling
- Packaging
- Published as the gitnexus npm package; supports npx and global npm installation
- Interfaces
- CLI, MCP tools, browser Web UI, and backend serve mode
- Local storage
- CLI uses persistent native LadybugDB; browser mode uses in-memory LadybugDB WASM
- Parsing
- CLI uses native Tree-sitter bindings; browser mode uses Tree-sitter WASM
- Testing structure
- Vitest unit and integration suites, Playwright browser tests, and Node server tests are present
Read from README.md, package.json, docs/plans/2026-08-04-gitnexus-plan-python-dotted-namespace-receiver.md, docker-server.mjs, eslint.config.mjs, docker-server.test.mjs, eval/__init__.py, eval/run_eval.py, eval/constants.py, eval/tool_registry.py, gitnexus/vitest.config.ts, gitnexus-web/vite.config.ts, gitnexus-web/vitest.config.ts, gitnexus-web/playwright.config.ts, eslint-rules/require-safe-parse.mjs.
What it can do
Create interactive knowledge graphs from codebases
Source code repository or codebase → Interactive knowledge graph visualization
Map code dependencies
Source code files → Dependency mapping and visualization
Trace call chains in code
Source code with function calls → Call chain mapping and flow visualization
Analyze execution flow paths
Source code files → Execution flow diagrams and analysis
Index codebases for searchable analysis
Source code repository → Indexed codebase with search capabilities
Provide architectural context to AI agents
Analyzed codebase knowledge graph → Structured architectural data for AI consumption
Enable code exploration through web interface
Indexed codebase → Interactive web-based code exploration interface
Integrate with AI agents via MCP protocol
Codebase analysis data → MCP-compatible data feed for AI agents
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.