Skip to content

Understand Anything

Egonex-AI/Understand-Anything

Understand Anything is a free, MIT-licensed plugin of 9 agent skills that turns a codebase, knowledge base or docs into a knowledge graph. It is made by Egonex and was created by Yuxiang Lin. It installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot and more than 10 other agents, and opens a browser dashboard to explore the graph.

By Egonex and Yuxiang LinMIT licenseWorks with 16 agentsResearch and data

Facts checked on against the sources listed at the end of this page.

GitHub stars
85,752
Skills in the pack
9
Agents supported
16named in the README
License
MIT
Last commit
14 hours ago9 October 2026
Instruction size
14.1k wordsacross all SKILL.md files

How to install Understand Anything

AgentCommandCopy
Claude Code
/plugin marketplace add Egonex-AI/Understand-Anything
/plugin install understand-anything
Codex
curl -fsSL https://raw.githubusercontent.com/Egonex-AI/Understand-Anything/main/install.sh | bash -s codex

On Windows, run iwr -useb https://raw.githubusercontent.com/Egonex-AI/Understand-Anything/main/install.ps1 | iex in PowerShell. Codex skills are invoked with $, for example $understand.

Cursor

No command. Cursor finds the plugin through .cursor-plugin/plugin.json when the repo is cloned. If it does not, open Cursor Settings, then Plugins, and paste https://github.com/Egonex-AI/Understand-Anything into the search field.

Gemini CLI
curl -fsSL https://raw.githubusercontent.com/Egonex-AI/Understand-Anything/main/install.sh | bash -s gemini
GitHub Copilot
copilot plugin install Egonex-AI/Understand-Anything:understand-anything-plugin

Copilot CLI command. In VS Code 1.108 or newer, the plugin is found through .copilot-plugin/plugin.json when the repo is cloned.

Commands for 1 more agent
Other agents
curl -fsSL https://raw.githubusercontent.com/Egonex-AI/Understand-Anything/main/install.sh | bash -s <platform>

Replace <platform> with the agent, for example opencode, openclaw, antigravity or kiro, as the README lists.

Commands as given in the README on 9 October 2026.

What Understand Anything runs on your machine

Hooks
Yes. hooks/hooks.json defines a PostToolUse hook on Bash that runs post-tool-use-auto-update.mjs and a SessionStart hook that tells the agent to refresh a stale graph when autoUpdate is true in the project config. The README says /understand --auto-update installs a post-commit hook.
Bundled scripts
Yes. install.sh and install.ps1 clone the repo to ~/.understand-anything/repo and symlink it for the chosen agent. The dashboard viewer runs through npx.
Network calls
Yes. The installers download from raw.githubusercontent.com and the viewer downloads a package from GitHub Releases. The README says the viewer serves the graph from local disk with no LLM calls. Analysis uses your agent's model and the README notes high token use. understand-figma calls api.figma.com.
API keys needed
Only for understand-figma, which needs a FIGMA_TOKEN environment variable. The other skills name none.

These are facts read from the repo, not a security rating. Read the files before you install any skill.

What is inside Understand Anything

SkillWhat it makes the agent do
understandAnalyzes the codebase and writes a knowledge graph to .ua/knowledge-graph.json. Re-runs update it incrementally.
understand-dashboardOpens the interactive graph dashboard in the browser.
understand-chatAnswers questions about the codebase using the graph.
understand-diffAnalyzes the impact of your current changes on the graph.
understand-explainExplains a specific file or function in depth.
understand-onboardGenerates an onboarding guide for new team members.
understand-domainExtracts business domains, flows and steps.
understand-knowledgeTurns a Karpathy-pattern LLM wiki into a graph.
understand-figmaAnalyzes a Figma file through the Figma REST API into a design knowledge graph.

A good fit for

  • Teams that onboard people into a large codebase and want a map to start from
  • Developers who inherit a project and need to see how its parts connect
  • Codebases where a change needs an impact check before it merges
  • People who keep a wiki or docs folder and want it as a searchable graph

Skip it if

  • Small projects where reading the code is quicker than building a graph
  • Token-sensitive setups, since the README notes significant token use for analysis
  • Teams that do not allow hooks that run after commands or at session start
  • Anyone who needs the graph to stay fresh without turning on auto-update

How Understand Anything works

The /understand skill scans a project in several phases and saves the result as a knowledge graph in the project's .ua folder. The other skills read that graph: a dashboard to browse it, a chat command to ask questions, a diff command to see what a change touches, and explain and onboard commands for new people.

The README suggests committing the .ua folder so a team shares one graph, except intermediate files and diff-overlay.json. It also mentions an optional local model setup through Ollama for privacy.

Works with
Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode, OpenClaw, Antigravity, Pi Agent, Vibe CLI, Hermes, Cline, KIMI CLI, Trae, Nanobot, Kiro

How much context Understand Anything uses

All 9 skills14,110 words across every SKILL.md file
Largest skillunderstand, 7,649 words
Loaded at startOnly each skill's name and description. The full text loads when a task needs that skill.

Word counts are measured from the repo every day. Reference files a skill loads on demand are not counted.

Is Understand Anything still maintained?

Last commit, 14 hours ago
Latest releasev2.9.0, 10 July 2026
Stars gained in 7 daysShown after 7 days of tracking
Stars gained in 30 daysShown after 30 days of tracking
Open issues and pull requests317
Forks7,215

Tools Understand Anything works with

Similar skill packs

Questions about Understand Anything

Is Understand Anything free?

Yes. It is open source under the MIT license. Analysis uses your agent's model, so it costs whatever tokens your plan charges.

Which agents does Understand Anything support?

The README gives install steps for Claude Code, Codex, Cursor, Gemini CLI and GitHub Copilot, plus OpenCode, OpenClaw, Antigravity, Pi Agent, Vibe CLI, Hermes, Cline, KIMI CLI, Trae, Nanobot and Kiro.

Where does it store the graph?

In .ua/knowledge-graph.json inside the analyzed project. Older projects keep using the .understand-anything folder.

Does the dashboard send my code anywhere?

The README says the viewer serves the graph read-only from local disk, with no LLM calls and no data leaving the machine.

Sources

Something wrong or out of date? Email [email protected] with a link to the source. Stats come from the public GitHub API and update every day.

For the maker

Made Understand Anything? Claim this page to show readers you check it.

Add this badge to your README. The number updates every day.

Understand Anything is listed on ScratchDB

[![Understand Anything on ScratchDB](https://scratchdb.com/badge/skills/understand-anything.svg)](https://scratchdb.com/skills/understand-anything/)

Sharing on X? This page shows a share card with the pack's live numbers.

Understand Anything agent skills: install, what is inside and supported agents | ScratchDB