finds.dev

Egonex-AI/Understand-Anything

Graphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. Works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more.

View on GitHubHomepage ↗

★ 88,083TypeScriptMITupdated Oct 2026

// the take

A Claude Code (and multi-platform CLI/IDE) plugin that runs a multi-agent pipeline over a codebase, combining tree-sitter static analysis with LLM summarization to build a browsable knowledge graph, dashboard, and guided tours. Aimed at developers onboarding to large unfamiliar codebases or teams wanting a shared, committable map of their architecture.

The tree-sitter/LLM split is the right call architecturally: imports, call sites, and inheritance come from deterministic parsing (reproducible, cheap to re-run), while the LLM is reserved for what parsers can't do — plain-English summaries and domain mapping. Incremental updates via fingerprint-based change detection mean only the first full run is expensive; subsequent runs re-analyze just the changed files. The graph is a portable JSON artifact — teammates can view it with a standalone npx viewer with no LLM calls, no API key, and nothing leaving the machine, which is a real win for onboarding docs-as-code. File analysis is parallelized with bounded concurrency and batching, which matters once you're pointing this at a few hundred thousand lines.

The first `/understand` run on a large repo can burn a serious number of tokens — the README itself tells you to budget for it or fall back to a local model, and there are no concrete cost numbers up front so you're finding out the hard way. It's shipped as an agent-CLI plugin/skill, not a library or service, so there's no straightforward way to run it headless in CI to keep a graph fresh as part of a pipeline. The semantic layer (summaries, layer/domain assignment) is entirely dependent on LLM output quality, and swapping to a local model for cost reasons is mentioned but with no stated accuracy tradeoff. The sheer volume of recently-dated internal design docs (theme system rewrites, language-agnostic overhaul, token reduction passes, Figma support) suggests the graph schema and pipeline internals are still moving fairly fast, which is worth knowing if you're committing `.ua/` output and expecting it to stay compatible release to release.

We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →

// at a glance

Stars
88,083
Forks
7,433
Language
TypeScript
License
MIT
Last commit
22 Oct 2026

// nearby

If Understand-Anything isn't quite it.

All repos →