A pre-indexed code knowledge graph that auto-syncs on every change and runs 100% locally. One MCP call returns entry points, related symbols, and snippets – fewer tokens, fewer tool calls, no data leaves your machine.
A pre-indexed code knowledge graph that auto-syncs on every change and runs 100% locally. One MCP call returns entry points, related symbols, and snippets - fewer tokens, fewer tool calls, no data leaves your machine.This is the skill's description: the text your AI reads to decide when the skill applies.
codegraph-local/SKILL.md
---
name: CodeGraph Local
description: A local, always-fresh knowledge graph of the codebase. When you need to orient - entry points, a symbol's source and dependents, call sites - ask CodeGraph in one call instead of reading many files, and trust the result without re-grepping.
---
# CodeGraph Local
CodeGraph Local keeps a pre-indexed SQLite knowledge graph of the project's
symbols and relationships, auto-synced on code changes by native OS file
watchers. Everything runs locally - no external APIs, no data leaves the
machine. Use it to orient fast and spend fewer tokens and tool calls.
## The four tools
- **explore** - answer a structural question ("how does auth flow work?") in a
single call: entry points, related symbols, and code snippets together.
- **node** - one symbol's source plus everything that depends on it.
- **search** - locate symbols by name across the codebase (FTS5 full-text).
- **callers** - find every call site of a function.
## How to use it well
- **Start with `explore`** for any "where do I begin" question - it builds the
context for you in one shot, instead of opening files one at a time.
- **Use `node` and `callers`** to scope a change before editing: see the source… load the full skill through Skill Me