Guide · Open source
How to use Understand-Anything: install, analyze and explore
Updated · written with AI from the project's own documentation and checked against our catalogue data
Understand-Anything is an open-source plugin that scans a project and builds an interactive knowledge graph of its files, functions, classes, and dependencies. It is aimed at developers who need to learn an unfamiliar codebase or see how modules depend on each other. By the end of this guide you will know how to install the plugin, run the analysis, open the dashboard, and ask questions about the code.
Understand AnythingEgonex-AI/Understand-Anything · ★ 89.7k · MIT Install options & downloads Quick start
- 1Add the marketplace. In Claude Code, run /plugin marketplace add Egonex-AI/Understand-Anything.
- 2Install the plugin. Run /plugin install understand-anything to add the plugin.
- 3Analyze the project. Run /understand in your project root to build the knowledge graph into .ua/knowledge-graph.json.
- 4Open the dashboard. Run /understand-dashboard to browse the graph visually.
- 5Ask and check changes. Use /understand-chat for questions and /understand-diff to see the impact of your current changes.
What you need before installing
Understand-Anything is built as an add-on for AI coding platforms. The main requirement listed in the project is Claude Code or another supported AI coding platform, such as Codex, Cursor, Copilot, or Gemini CLI.
- Claude Code, or another supported AI coding platform
- A project (codebase) you want to analyze
- Enough token budget for the first run, which can be large on big projects
Install the plugin in Claude Code
Installation is done with two slash commands inside Claude Code. First add the plugin marketplace, then install the plugin.
/plugin marketplace add Egonex-AI/Understand-Anything
/plugin install understand-anythingAnalyze your codebase with /understand
Run /understand in your project. A multi-agent pipeline scans the project, extracts every file, function, class, and dependency, and saves a knowledge graph to .ua/knowledge-graph.json. Projects that already have a .understand-anything/ directory keep using it as the data directory.
The first run on a large project can use a significant number of tokens. Later runs are incremental by default, so only changed files are re-analyzed.
- Use --language to set output language: en (default), zh, zh-TW, ja, ko, ru, vi
- Language choice is saved to .ua/config.json and reused on later runs
- Scope to a subdirectory on large monorepos by passing a path, for example src/frontend
/understand
/understand --language zh
/understand src/frontendExplore the graph in the dashboard
Run /understand-dashboard to open an interactive web dashboard. The graph is color-coded by architectural layer, searchable, and clickable. Selecting a node shows its code, relationships, and a plain-English explanation.
The structural graph is one view. The domain view, opened from the dashboard, shows how code maps to business processes, including domains, flows, and steps.
- Guided tours walk through the architecture in dependency order
- Fuzzy and semantic search, for example asking which parts handle auth
- Layer grouping such as API, Service, Data, UI, and Utility
/understand-dashboardAsk questions and check changes
Several commands cover follow-up work. Use /understand-chat to ask a question about the codebase, /understand-explain to look at one file or function, and /understand-diff to see which parts of the system your current changes might affect before you commit.
Other commands generate an onboarding guide, extract domain knowledge, or analyze a Karpathy-pattern LLM wiki. To keep the graph current automatically, the README describes running /understand --auto-update through a post-commit hook.
/understand-chat How does the payment flow work?
/understand-explain src/auth/login.ts
/understand-diff
/understand-onboard
/understand-domain
/understand-knowledge ~/path/to/wiki
/understand --auto-updateUse a local model for privacy
For privacy or enterprise setups, the README suggests pointing your platform at a local model provider such as Ollama. Follow the Ollama integration guide for your platform to change the model provider. This can also reduce the token load for initialization, according to the project's advice.
What people use it for
- Onboarding onto a new team's codebase by following guided tours in dependency order
- Seeing how files and modules depend on each other as a visual map
- Checking what your current changes might affect before committing with /understand-diff
- Viewing a Karpathy-pattern LLM wiki as a linked, navigable graph with /understand-knowledge
Frequently asked questions
Is Understand-Anything free?
The project is released under the MIT licence, and its source is on GitHub. The README does not state pricing for the AI platforms it runs on, and the first analysis of a large project can use many tokens, so check your platform's usage terms.
What do I need to install Understand-Anything?
The listed prerequisite is Claude Code or another supported AI coding platform. You install it with two commands: /plugin marketplace add Egonex-AI/Understand-Anything and /plugin install understand-anything.
Does Understand-Anything work with Cursor or Copilot?
The project says it works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more. The README also lists Copilot CLI, OpenCode, Vibe CLI, and Trae sections. Check the platform section of the README for setup details on your tool.
Why is the first run slow or expensive?
The initial /understand analyzes the whole codebase, which can consume a significant number of tokens on large projects. Later runs are incremental and only re-analyze changed files. Scoping to a subdirectory, such as src/frontend, can also reduce the work.
Can I use a local model instead of a cloud model?
Yes, the README suggests pointing your platform at a local model provider such as Ollama for privacy or enterprise setups. Follow the Ollama integration guide to change the model provider.
Is the analysis deterministic?
No. The project relies on LLM agents for much of its work, so it is not a fully deterministic analysis tool. If you need a non-AI, repeatable analyzer, pick a different kind of tool.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Understand-Anything suits developers joining a new team, or anyone who wants a visual map of how files and modules depend on each other, and who already uses a supported AI coding platform. It is less suitable if you have strict token or cost budgets, need fully deterministic analysis, or do not use one of the supported platforms.
Install options & downloads