Repo Year in Review

Egonex-AI/Understand-Anything

A Claude Code plugin that analyzes a codebase or wiki with a multi-agent pipeline and builds an interactive knowledge graph and dashboard to explore, search, and query it.

  • 85.8k GitHub stars
  • TypeScript
  • ⚖️ MIT
View on GitHub (opens in a new tab) Setup guide
Egonex-AI/Understand-Anything preview image

What it is

Understand Anything is a Claude Code plugin that scans a project with a multi-agent pipeline and builds a knowledge graph of every file, function, class, and dependency. It combines deterministic Tree-sitter parsing with LLM-generated summaries, layers, and guided tours. An interactive dashboard lets you explore the graph, search it, and ask questions about the code. It also works on Karpathy-pattern LLM wiki knowledge bases and runs on many other AI coding platforms.

Who it's for

  • Developers who just joined a team and need to learn a large codebase
  • Teams that want to onboard people or review PRs using a shared, committed knowledge graph
  • Users of AI coding platforms such as Claude Code, Codex, Cursor, Copilot, and Gemini CLI
  • People who want to turn a Karpathy-pattern LLM wiki into a navigable graph

Requirements

Requirements

  • Claude Code, or another supported AI coding platform (Codex, Cursor, Copilot, Gemini CLI, OpenCode, Vibe CLI, Trae, Kiro, etc.)
  • Node.js (>= 18) to view the dashboard without Claude Code
  • Enough LLM token budget for the first full /understand run; the README recommends a token plan/subscription or a local model for initialization on large projects
  • VS Code with GitHub Copilot v1.108+ for auto-discovery in VS Code

Setup

  1. Install the plugin in Claude Code

    Add the marketplace and install the plugin.

    bash
    /plugin marketplace add Egonex-AI/Understand-Anything
/plugin install understand-anything

  2. Analyze your codebase

    Runs the multi-agent pipeline and saves the knowledge graph to .ua/knowledge-graph.json.

    bash
    /understand

  3. Open the dashboard

    Opens the interactive web dashboard showing your codebase as a graph.

    bash
    /understand-dashboard

  4. One-line install for other platforms (macOS / Linux)

    Installs for Codex, OpenCode, Gemini CLI, and other listed platforms. Pass the platform name to skip the prompt. Restart your CLI/IDE afterwards.

    bash
    curl -fsSL https://raw.githubusercontent.com/Egonex-AI/Understand-Anything/main/install.sh | bash
# or skip the prompt by passing the platform:
curl -fsSL https://raw.githubusercontent.com/Egonex-AI/Understand-Anything/main/install.sh | bash -s codex

  5. One-line install for other platforms (Windows)

    PowerShell installer.

    powershell
    iwr -useb https://raw.githubusercontent.com/Egonex-AI/Understand-Anything/main/install.ps1 | iex

  6. Install on Copilot CLI

    Install via the Copilot CLI plugin command.

    bash
    copilot plugin install Egonex-AI/Understand-Anything:understand-anything-plugin

Examples

Ask a question about the codebase

Prompt
prompt
/understand-chat How does the payment flow work?

Expected output: Queries the generated knowledge graph in natural language.

Explain a specific file

bash
bash
/understand-explain src/auth/login.ts

What it does: Deep-dives into a specific file or function.

Generate output in another language

bash
bash
/understand --language zh

What it does: Generates node descriptions and dashboard UI in Chinese; supported languages are en (default), zh, zh-TW, ja, ko, ru, vi.

Keep the graph updated on every commit

bash
bash
/understand --auto-update

What it does: Enables a post-commit hook that incrementally patches the graph so each commit lands with a matching graph.

View a committed graph without Claude Code

bash
bash
npx https://github.com/Egonex-AI/Understand-Anything/releases/latest/download/understand-anything-viewer.tgz /path/to/analyzed/project

What it does: Opens the full interactive dashboard from a committed graph using only Node.js (>= 18), with no LLM or API key.

Pros & cons

Pros

  • Pro:Hybrid design: deterministic Tree-sitter parsing gives reproducible structural edges, while the LLM adds summaries, layers, and tours
  • Pro:Incremental re-analysis of only changed files, plus an optional post-commit auto-update hook
  • Pro:Graph is plain JSON that can be committed and viewed by teammates without Claude Code, an LLM, or an API key
  • Pro:Broad platform support, including Claude Code, Codex, Cursor, Copilot, Gemini CLI, and many more

Cons

  • Con:The initial /understand run analyzes the whole codebase and can consume a significant number of tokens on large projects
  • Con:Invocation differs by platform (Codex uses $understand instead of /understand), and some platforms may require asking in plain language

Images

Egonex-AI/Understand-Anything preview image
Egonex-AI/Understand-Anything preview image · Egonex-AI/Understand-Anything on GitHub · MIT
Egonex-AI/Understand-Anything preview image