GitNexus
A tool that pre-reads your entire codebase and builds a map of how every file and function connects, so an AI coding assistant can find the right piece of code instantly instead of having to search through your whole project every time.
🔗 Visit GitNexus
Description
AI coding assistants often struggle with large codebases because they can only read a limited amount of code at once, and don't inherently know how files, functions and modules relate to each other — so they either miss important context or waste effort re-reading the same files. GitNexus solves this by doing that mapping work up front, once, and letting the AI query the result instead of re-discovering it every session.
GitNexus is a free, open-source (PolyForm Noncommercial 1.0.0, 44,600+ GitHub stars) knowledge-graph engine that uses compiler-grade static analysis to index a codebase into a queryable dependency and execution-flow graph, all processed locally for privacy. It supports 14+ languages (TypeScript, Python, Java, Go, Rust, C# and more), integrates with Claude Code, Cursor, Codex and Windsurf via MCP, supports multi-repo analysis with cross-repo edge detection, and uses Leiden clustering to automatically identify module and system structure. On the SWE-bench Verified benchmark it's reported to achieve a 30% token reduction while solving 116/162 problems. An enterprise SaaS/self-hosted option is available through its backing company, Akon Labs (Y Combinator S26 batch).
💬 Our review
The short version: GitNexus is one of a small handful of tools trying to give AI coding agents real structural understanding of a codebase instead of just raw text search, and its reported 30% token reduction on a real benchmark is a concrete, checkable claim rather than vague marketing.
The closest comparison in this catalog is Code-Review-Graph, which builds a similar local-first code intelligence graph for MCP and CLI use — the two overlap conceptually but GitNexus leans harder into compiler-grade static analysis and multi-repo cross-referencing with automatic clustering (Leiden), while positioning itself more explicitly around benchmark-proven token savings for agent workflows. Both are free and open source at their core, so the practical choice comes down to which one's language coverage and integration (Claude Code, Cursor, Codex, Windsurf for GitNexus) fits your stack — trying both on a real project is cheap since neither costs money to start. Being backed by a YC-batch company (Akon Labs) offering an enterprise tier is a reasonable signal of staying power for a project this new, though the enterprise pricing itself isn't public yet.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Version open-source gratuite (licence PolyForm Noncommercial). Offre entreprise SaaS/self-hosted via Akon Labs, tarifs sur devis.
Pros
Analyse statique de niveau compilateur, traitement 100% local
Réduction de tokens mesurée (-30% sur SWE-bench Verified)
Support de 14+ langages et analyse multi-dépôts
Intégration MCP avec Claude Code, Cursor, Codex, Windsurf
Cons
Chevauche conceptuellement Code-Review-Graph, choix à tester au cas par cas
Tarification entreprise non publique
Licence PolyForm Noncommercial : usage commercial nécessite l'offre payante