Understand-Anything: a Claude Code plugin that turns codebases into interactive knowledge graphs

Understand-Anything ships as a Claude Code plugin. Its multi-agent pipeline scans a project, extracts every file, function, class, and dependency, and saves the resulting knowledge graph to .ua/knowledge-graph.json. Tree-sitter parses source code for structural facts like imports, exports, and function/class definitions, while the LLM adds plain-language summaries, architecture layers, business-domain mapping, and guided tours. Features include fuzzy and semantic search, diff impact analysis, architecture-layer grouping and visualization, a Persona-adaptive UI that adjusts detail for junior devs, PMs, or power users, and language-concept explanations across 12 programming paradigms. The first /understand run analyzes the whole codebase, which can burn a lot of tokens on large projects, so the team recommends initializing on a subscription plan or a local model such as Ollama; later runs are incremental by default, re-analyzing only changed files. The graph is JSON and can be committed so teammates skip the analysis pipeline — use git-lfs for files over 10 MB. Beyond the native Claude Code plugin, it supports Codex, Cursor, Copilot, Gemini CLI, Cline, Kimi CLI, and others (note Codex uses the $understand prefix instead of a slash command). For projects with a committed graph, teammates need no Claude Code, LLM, or API key — just Node.js 18+ to open a read-only dashboard via npx. This workflow is directly useful for new members ramping up on large codebases, engineers doing PR reviews, and teams writing onboarding docs.