Ast Outline — AI skill for Claude Code

ast-outline

**English** · [Русский](./README.ru.md) · [简体中文](./README.zh-CN.md)

Your AI coding agent reads whole files to understand code. **`ast-outline` lets it pull exactly what it needs** — a whole-repo skeleton, a single file's outline, or one symbol's body — instead of reading files in full. One stateless, tree-sitter-based CLI: map a repo, outline a file, show a symbol, structural-grep for usages, trace imports. Cheaper, faster agent runs that don't drown in large codebases.

[](https://github.com/ast-outline/ast-outline/actions/workflows/tests.yml) [](./LICENSE) [](./LICENSE-DOCS) [](https://pypi.org/project/ast-outline/) 

📖 **Full documentation:**

---

Why

LLM coding agents (Claude Code, Cursor agent mode, Aider, Codex CLI, Gemini CLI, Copilot Chat) explore codebases by **reading files directly**. Reliable, but wasteful in two ways: a 1200-line file costs 1200 lines of context just to answer *"what methods are in here?"* — and once the noise is in context, the agent has to wade through it to find the part that actually matters. Token cost goes up, comprehension goes down.

`ast-outline` is a pre-reading layer. The agent calls it first, gets the file's shape in 60–100 lines, and only opens the bodies it actually needs. The win is double: **fewer tokens** in context, and **sharper comprehension** — less noise to filter through means the agent locks onto the relevant code faster and answers with less drift.

**Before:**

Agent: Read Player.cs              # 1200 lines, just to see what's here
Agent: Read Enemy.cs               #  800 lines