codegraph

CodeGraph is an open-source, 100% local code knowledge graph designed to supercharge AI coding agents including Claude Code, Cursor, Codex CLI, opencode, Hermes Agent, Gemini CLI, Antigravity IDE, Kiro, and GitHub Copilot. It builds a pre-indexed semantic graph of your entire codebase — every symbol, call edge, and dependency — and auto-syncs on every file change so the index is never stale. Instead of agents crawling files via grep and Read tool calls, CodeGraph gives them the exact code they need in one call. Key capabilities: - **Native Rust Kernel**: Parsing engine written in Rust for 20+ languages (TypeScript, JavaScript, Python, Go, Rust, Java, C/C++, Swift, Kotlin, and more). Graphs are verified byte-for-byte identical to the reference engine. - **Adapts to Your Machine**: Worker pools and caches are sized from actual system resources — container-aware core counts and honest RAM measurements. Works on both workstations and low-spec VPS. - **Surgical Context**: One tool call returns entry points, related symbols, and code snippets, eliminating file-by-file exploration. - **Auto-Sync**: Watches the project and updates the graph on every file change (~0.3s incremental sync). - **Framework-Aware Routes**: Understands routing in web frameworks. - **Mixed Platform Support**: Handles iOS/React Native/Expo bridging. - **Browser UI**: Visualize the graph at http://127.0.0.1:4747 showing callers, symbols, and callees. - **MCP Server**: Integrates via Model Context Protocol for supported agents. Benchmark results (measured 2026-08 across 7 open-source repos with Claude Opus 4.8): CodeGraph reduced tool calls by 88%, tokens by 62%, and cost by 44% compared to file-reading agents, with file reads cut to zero on all tested repos. Installation: standalone binary (no Node.js required) via install script, or npm global package. Wired up per-project with `codegraph init` and connected to agents with `codegraph install`.