GitNexus: Map Codebases for AI-Assisted Analysis

Overview

GitNexus builds a knowledge graph of a repository, covering code relationships such as dependencies, call chains, functional clusters, and execution flows. Its CLI and Model Context Protocol (MCP) server let developers and compatible AI coding agents query that structure instead of relying only on file-by-file search or generated descriptions.

Use it when understanding a codebase or checking the effects of a change is difficult from local context alone. A browser-based UI is also available for visual exploration and chat; local CLI indexing is the recommended route for ongoing development and larger repositories.

Key Features

• Indexes source code into a graph of symbols, dependencies, clusters, and execution flows.
• Provides MCP tools for code search, symbol context, dependency impact, path tracing, and change analysis.
• Supports AI editor integrations through MCP, with skills and hooks available for selected editors.
• Offers a CLI for indexing and querying repositories, plus a local HTTP server that connects the web UI to local indexes.
• Includes a browser UI for graph exploration and repository chat, with browser-only indexing limited by available memory.
• Supports repository groups for cross-repository contract discovery and impact analysis.

Use Cases

• A developer joining an unfamiliar project can explore its functional areas, entry points, and execution flows before making changes.
• A team using AI coding agents can provide graph-backed context to help agents account for callers and dependencies when editing code.
• Maintainers can inspect the likely blast radius of a change before a commit or refactor.
• Engineers working across related services can group repositories to inspect contracts and cross-repository links.

Getting Started

From the root of a Git repository, run:

npx gitnexus analyze
npx gitnexus setup

See the README for editor configuration, deployment options, and additional CLI details.

Alternatives

• codegraph: CodeGraph builds a local code-structure graph for MCP queries, while GitNexus focuses on indexing codebases and exposing relationships through CLI and MCP tools.
• codebase-memory-mcp: codebase-memory-mcp is a persistent local MCP server for code-structure queries, while GitNexus also offers a CLI for tracing flows and change impact.
• GitNexus: This GitNexus runs in the browser on GitHub repositories or ZIP files and includes an interactive graph and Graph RAG Agent, rather than focusing on CLI and MCP tools.
• Understand-Anything: Understand-Anything also graphs code, but extends analysis to documentation and knowledge bases and emphasizes interactive exploration through AI coding platforms.

Considerations

• The GitHub metadata reports the license as NOASSERTION, while the README displays a PolyForm Noncommercial license. Review the repository's license terms before adopting it, especially for commercial use.
• Local indexing depends on native database and parser components. The README documents installation caveats, including npm compatibility issues and optional C++ toolchain requirements for some grammars.
• The browser-only UI is constrained by browser memory; the README estimates this mode at roughly 5,000 files. For larger repositories, use local CLI indexing or backend mode.
• Some editor integrations provide MCP only, while hooks and skill support vary by editor. Check the README for the current integration details.