Retrieve The Forgotten Memory

11 tasks, 3 repos of varying size, 4 conditions (A = standard prompt with file paths; B = discovery, no paths; C = RTFM FTS; D = RTFM hybrid), 3 runs each.

On a single smaller-scope run (test_stub_generatoron metaflow), RTFM cut agent time by−37 %vs the no-paths baseline. On the larger repos, the tasks themselves were too hard for Sonnet 4 to resolve inside a 20-minute timeout regardless of retrieval.

- Single model (Sonnet 4), single agent (Claude Code). Not statistically bullet-proof.
- On small repos (< 1k files),grepis enough and RTFM adds overhead.
- FeatureBench measurescode modification, notinformation retrieval. It's the wrong benchmark for a retrieval tool — I'm running against it because it's what exists. Better-suited benchmarks (RepoQA, SWE-QA, LocAgent) are on the roadmap.

RTFM measurably wins when the bottleneck is"find the right paragraph in a 2,000-file corpus". It doesn't magically make unsolvable tasks solvable. The model still has to do the work — RTFM just makes sure it has the right context to do it with.

RTFM works anywhere your project isn't just code:

- LegalTech— Code + tax law + regulatory specs. Ships with Legifrance XML and BOFiP parsers.
- Research— Code + LaTeX papers + datasets. Ships with LaTeX and PDF parsers.
- FinTech— Code + financial regulations + XBRL reports. Write an XBRL parser in 50 lines.
- HealthTech— Code + medical records (HL7/FHIR) + clinical guidelines.
- Solo devs with big projects— Stop watching your agent grep the same 8,000 files every session.
- Obsidian / PKM users— Make your vault actually searchable by your AI.
- Any regulated industry— If your project mixes code with domain documents, RTFM is for you.

- FTS5 full-text search— instant, zero-config, works out of the box
- Semantic search— optional embeddings (FastEmbed/ONNX, no GPU needed)
- Hybrid mode— combine both, rank by relevance score
- Metadata-first— results return file paths + scores (~300 tokens), not content dumps
- Progressive disclosure— agent expands only the chunks it actually needs
- Knowledge graph— wikilinks + Python imports resolved as graph edges, hub detection, centrality ranking

- 22 parsers built-in— Markdown, Python (AST), LaTeX, YAML, JSON, TOML, Shell, PDF, XML, HTML, SQLite, Jupyter, CSV/TSV, XLSX, EPUB, MOBI/AZW, FB2, DJVU, DOCX, ODT, RTF, plain text
- Extensible— add any format in ~50 lines of Python
- Auto-sync hooks— index stays fresh every prompt, zero manual work
- Incremental— only re-indexes what changed

- Native Claude Code plugin—/plugin install rtfm@roomi-fields/rtfm, auto-init per project
- Pure-Python MCP server— 0 external deps, nomcpSDK /pydantic/ native binaries
- Cross-platform— Linux, macOS, Windows, WSL (only requires Python ≥ 3.10 on PATH)
- 13 MCP tools— search, context, expand, graph, history, sync, tags, ...
- Manual install fallback—pip install rtfm-aifor Cursor, Codex, Claude Desktop chat, any other MCP client
- CLI + Python API— scriptable for pipelines
- Non-invasive— doesn't touch your code, doesn't replace your editor

Need to index a format nobody supports? Write a parser in ~50 lines.

# Search rtfm search "authentication flow" rtfm search "article 39" --corpus cgi --limit 5 # Sync rtfm sync # All registered sources rtfm sync /path/to/docs --corpus docs # Specific directory rtfm sync . --force # Force re-index # Source management rtfm add /path/to/docs --corpus docs --extensions md,pdf rtfm sources # Obsidian vault rtfm vault # Initialize for cwd vault rtfm vault /path/to/vault # Specific vault rtfm vault --regenerate # Regenerate _rtfm/ files # Cross-project Claude memory rtfm memory # Manual snapshot rtfm memory --install-hook # Auto-snapshot on SessionEnd # Status & info rtfm status rtfm books rtfm tags rtfm history path/to/file.md # Memory version history # Semantic search rtfm embed # Generate embeddings (one-time) rtfm semantic-search "tax deductions" --hybrid # MCP server rtfm serve

from rtfm import Library lib = Library("my_library.db") # Index stats = lib.ingest("documents/article.md", corpus="docs") result = lib.sync(".", corpus="my-project") # SyncResult(+3 ~1 -0 =42) # Search results = lib.search("depreciation", limit=10, corpus="cgi") results = lib.hybrid_search("amortissement fiscal", limit=10) # Export for LLM prompt_context = results.to_prompt(max_chars=8000) lib.close()

RTFM isn't a task manager. It's not an agent framework. It's the knowledge layer your agent needs underneath whatever you're already using.

┌─────────────────────────────────┐ │ GSD / Taskmaster / Claude Flow │ ← Orchestration ├─────────────────────────────────┤ │ RTFM │ ← Knowledge (you are here) ├─────────────────────────────────┤ │ Claude Code │ ← Execution └─────────────────────────────────┘

Without RTFM, your orchestrator drives an agent that hallucinates. With RTFM, the agent knows what it's building on.

MIT — use it, fork it, extend it, ship it.

Code indexers see your code. RTFM sees everything.

This is a web browser that enables your coding agent, such as Claude Code, to visit websites on your behalf and assist you in identifying bugs or creating UI test cases.

A Retrieval-Augmented Generation (RAG) server for document processing, vector storage, and intelligent Q&A, powered by the Model Context Protocol.

Local code analysis MCP server with 25+ tools: semantic search, call graph tracing, dependency analysis, and symbol navigation. Built with Tree-sitter and CozoDB. Supports Go, Python, JS, TS.

A local-first code indexer that enhances LLMs with deep code understanding. It integrates with AI assistants via the Model Context Protocol (MCP) and supports AI-powered semantic search.

A knowledge management tool for code repositories using vector embeddings, powered by a local Ollama service.

A server for managing structured project context using SQLite, with support for vector embeddings for semantic search and Retrieval Augmented Generation (RAG).