`, `&` in content |\n| Binary file marking | `[binary]` tag in directory tree | Listed but not marked |\n| Directory tree | Tree characters (`โโโ` `โโโ` `โ`) | Flat indentation |\n| .gitignore in output | Included | Omitted |\n\nThe key issue: repomix v1.12 doesn't XML-escape file content, so files containing `<`, `>`, or `&` produce **invalid XML** that breaks parsers. claudemem handles this correctly.\n\n## Symbol graph & code analysis\n\nBeyond semantic search, claudemem builds a **symbol graph** with PageRank scores. This enables powerful code analysis:\n\n### Dead code detection\n```bash\nclaudemem dead-code\n# Finds: symbols with zero callers + low PageRank + not exported\n# Great for: cleaning up unused code\n```\n\n### Test coverage gaps\n```bash\nclaudemem test-gaps\n# Finds: high-PageRank symbols not called by any test file\n# Great for: prioritizing what to test next\n```\n\n### Change impact analysis\n```bash\nclaudemem impact FileTracker\n# Shows: all transitive callers, grouped by file\n# Great for: understanding blast radius before refactoring\n```\n\n### Keep index fresh\n```bash\n# Option 1: Watch mode (daemon)\nclaudemem watch\n\n# Option 2: Git hook (auto-index after commits)\nclaudemem hooks install\n```\n\n## Documentation indexing\n\nclaudemem can automatically fetch and index documentation for your project dependencies. This gives you semantic search across both your code AND the frameworks you use.\n\n### How it works\n\n1. **Detects dependencies** from `package.json`, `requirements.txt`, `go.mod`, `Cargo.toml`\n2. **Fetches docs** using a provider hierarchy with automatic fallback:\n - **Context7** โ 6000+ libraries, versioned API docs & code examples (requires free API key)\n - **llms.txt** โ Official AI-optimized docs from framework sites (Vue, Nuxt, Langchain, etc.)\n - **DevDocs** โ Consistent offline documentation for 100+ languages\n3. **Chunks & indexes** documentation alongside your code\n4. **Search everything** with natural language queries\n\n### Setup\n\n```bash\nclaudemem init # prompts to enable docs & configure Context7\n```\n\nOr configure manually:\n```bash\nexport CONTEXT7_API_KEY=your-key # get free key at https://context7.com/dashboard\n```\n\n### Commands\n\n```bash\nclaudemem docs fetch # fetch docs for all detected dependencies\nclaudemem docs fetch react vue # fetch specific libraries\nclaudemem docs status # show indexed docs & cache state\nclaudemem docs clear # clear cached documentation\n```\n\n### What gets indexed\n\n| Source | Best For | Coverage | Auth Required |\n|--------|----------|----------|---------------|\n| **Context7** | Code examples, API reference | 6000+ libs | Free API key |\n| **llms.txt** | Official structured docs | 500+ sites | None |\n| **DevDocs** | Offline fallback | 100+ langs | None |\n\n### Configuration\n\nIn `~/.claudemem/config.json`:\n```json\n{\n \"docs\": {\n \"enabled\": true,\n \"providers\": [\"context7\", \"llms_txt\", \"devdocs\"],\n \"cacheTTL\": 24,\n \"maxPagesPerLibrary\": 10,\n \"excludeLibraries\": [\"lodash\"]\n }\n}\n```\n\nEnvironment variables:\n- `CONTEXT7_API_KEY` โ Context7 API key (optional but recommended)\n- `CLAUDEMEM_DOCS_ENABLED` โ disable docs entirely (`false`)\n\n## Supported languages\n\nTypeScript, JavaScript, Python, Go, Rust, C, C++, Java.\n\nIf your language isn't here, it falls back to line-based chunking. Works, but not as clean.\n\n## CLI reference\n\n### Basic commands\n```\nclaudemem init # setup wizard\nclaudemem index [path] # index codebase\nclaudemem search
# search (auto-reindexes changed files)\nclaudemem status # what's indexed\nclaudemem clear # nuke the index\nclaudemem models # list embedding models\nclaudemem benchmark # benchmark embedding models\nclaudemem --mcp # run as MCP server\n```\n\n### Symbol graph commands (for AI agents)\n```\nclaudemem map [query] # repo structure with PageRank scores\nclaudemem symbol # find symbol definition\nclaudemem callers # what calls this symbol?\nclaudemem callees # what does this symbol call?\nclaudemem context # symbol + callers + callees\n```\n\n### Code analysis commands\n```\nclaudemem dead-code # find potentially dead code (zero callers + low PageRank)\nclaudemem test-gaps # find important code without test coverage\nclaudemem impact # analyze change impact (transitive callers)\n```\n\n### Pack commands\n```\nclaudemem pack [path] # pack codebase to XML (default: -pack.xml)\nclaudemem pack --format md # markdown format\nclaudemem pack --stdout # write to stdout\nclaudemem pack --include \"src/**\" --exclude \"*.test.ts\"\n```\n\n### Pack flags\n```\n-o, --output # output file path\n--format # output format (default: xml)\n--stdout # write to stdout instead of file\n--include # glob pattern to include (repeatable)\n--exclude # additional exclusion pattern (repeatable)\n--no-gitignore # don't use .gitignore patterns\n--max-file-size # max file size (default: 1048576)\n--tokens # show token count report\n```\n\n### Developer experience\n```\nclaudemem watch # auto-reindex on file changes (daemon mode)\nclaudemem hooks install # install git post-commit hook for auto-indexing\nclaudemem hooks uninstall # remove the hook\nclaudemem hooks status # check if hook is installed\n```\n\n### IDE integrations\n```\nclaudemem install opencode # install OpenCode plugins (suggestion + tools)\nclaudemem install opencode --type tools # install tools plugin only\nclaudemem install opencode status # check installation status\nclaudemem install opencode uninstall # remove plugins\n```\n\n### Documentation commands\n```\nclaudemem docs fetch # fetch docs for all detected dependencies\nclaudemem docs fetch # fetch docs for specific library\nclaudemem docs status # show indexed docs and providers\nclaudemem docs clear # clear cached documentation\n```\n\n### Search flags\n```\n-n, --limit # max results (default: 10)\n-l, --language # filter by language\n-y, --yes # auto-create index without asking\n--no-reindex # skip auto-reindex\n```\n\n### Code analysis flags\n```\n--max-pagerank # dead-code threshold (default: 0.001)\n--min-pagerank # test-gaps threshold (default: 0.01)\n--max-depth # impact analysis depth (default: 10)\n--include-exported # include exported symbols in dead-code scan\n--agent # agent mode: no logo, compact output (for AI tools)\n```\n\n## Config\n\nEnv vars:\n- `OPENROUTER_API_KEY` โ for OpenRouter provider\n- `CLAUDEMEM_MODEL` โ override embedding model\n- `CONTEXT7_API_KEY` โ for documentation fetching (optional)\n\nFiles:\n- `~/.claudemem/config.json` โ global config (provider, model, docs settings)\n- `.claudemem/` โ project index (add to .gitignore)\n\n## Limitations\n\n- First index takes a minute on large codebases\n- Ollama is slower than cloud (runs locally, no batching)\n- Embedding quality depends on the model you pick\n- Not magic โ sometimes grep is still faster for exact strings\n\n## License\n\nMIT\n\n---\n\n[GitHub](https://github.com/MadAppGang/claudemem) ยท [npm](https://www.npmjs.com/package/claude-codemem) ยท [OpenRouter](https://openrouter.ai)\n","readmeFilename":"README.md"}![\"CLAUDEMEM\"](\"assets/logo.png\"){kind=logo}
\n![\"npm](\"https://img.shields.io/npm/v/claude-codemem.svg\")
\n ![\"License:](\"https://img.shields.io/badge/License-MIT-yellow.svg\"){kind=icon}
\n ![\"GitHub](\"https://img.shields.io/github/stars/MadAppGang/claudemem?style=social\")
![\"Embedding](\"assets/benchmark.png\")
\n