Where the Graphifycommunity talks.

An 82K-star GitHub repo is built around one painfully obvious idea: Your coding agent should map the codebase once, not grep it forever. Graphify turns an entire project into a queryable knowledge graph. Functions, classes, files, SQL schemas, infrastructure, docs, PDFs, images and videos become connected nodes that an agent can traverse instead of repeatedly opening files and reconstructing the architecture. So instead of: → search for authentication → open twelve files → follow imports manually → lose the trail as the context fills up The agent can ask: > What connects authentication to the database? > Trace the path from UserService to DatabasePool. > Explain RateLimiter. > Which concepts does everything flow through? Graphify returns the relevant subgraph and the path connecting the concepts, not another list of keyword matches. For source code, this is not RAG: → No embeddings → No vector database → No LLM required → Code is parsed locally using tree-sitter → Calls, imports and inheritance become graph edges Every relationship is also marked as EXTRACTED, INFERRED, or AMBIGUOUS, so the agent can distinguish what exists explicitly in the source from what Graphify resolved or guessed. The cleverest part is what happens next. Graphify can install hooks or persistent instructions for Claude Code, Codex, Cursor, Gemini CLI, Copilot and 20+ other assistants. Before the agent starts blindly grepping or reading files one by one, it is nudged to query the existing graph first. The graph can be committed to Git, automatically rebuilt after commits, shared across the team and exposed through MCP. Long context windows help agents read more code. A persistent knowledge graph helps them know where to look. The next improvement in coding agents may not come from stuffing more files into the prompt. It may come from making them stop rereading the repository. Here's the GitHub Repo: https://t.co/4ify3X8urp

Graphify Labs is the first YC company to hit 100K github stars during the batch.

After a weekend of using it on MemMachine, I’m not going back. We’re seeing 79x token reductions, and zero vector database needed.

Someone just built Karpathy's second brain, but for code. graphify turns files, PDFs, and screenshots into a knowledge graph you can query. 100% Open source Code in comments

THIS CLAUDE CODE SKILL MAKES CLAUDE 71.5X MORE EFFICIENT AT UNDERSTANDING YOUR CODEBASE it's called /graphify and it went viral in 26 days instead of claude guessing what your project looks like or wasting tokens exploring files one by one, the skill maps every function, every dependency, and every connection upfront claude gets full context immediately here's how it works: it extracts structure from code files locally with no AI calls. transcribes any video or audio locally with whisper. then uses claude to pull concepts and relationships from docs, papers, and images with parallel subagents everything merges into one graph, gets clustered, and exports as interactive HTML, queryable JSON, and a plain language report it handles: > code in 25 languages via tree-sitter AST > markdown, HTML, YAML, config files > PDFs with citation mining > images via claude vision (screenshots, diagrams, any language) > video and audio transcribed locally with whisper > youtube URLs downloaded and transcribed automatically > SQL files extracted deterministically (tables, views, foreign keys, joins) > docx and xlsx converted to markdown then extracted every relationship is tagged EXTRACTED (found in source), INFERRED (reasonable guess with a confidence score), or AMBIGUOUS (flagged for review). you always know what was found vs what was guessed people aren't just using it for code. they're using it for PhD research papers, obsidian vaults, personal databases, meeting transcripts, legal docs. anything with structure that needs to be understood fast works with claude code, codex, cursor, gemini CLI, github copilot, vs code, aider, kiro, and basically every AI coding tool that exists one user said GPT-5.5 + graphify is letting him push features faster than anything he's ever used set up post-commit hooks and the graph rebuilds automatically every time you push code. no manual reruns the token savings are insane. 71.5x fewer tokens per query vs reading raw files on a mixed corpus of code, papers, and images 40k github stars and 450k downloads one slash command turned claude code into a context machine

Instead of Claude Code re-reading every file, it queries the graph — which is persistent across sessions and costs a fraction of the tokens.

Someone built Andrej Karpathy's dream tool 48 hours after he asked for it. Graphify is an open-source tool for Claude Code. Point it at any folder and one command builds a knowledge graph. It reads your code, docs, PDFs, and images. No vector database. No config files. What comes out the other side: > Navigable graph of every concept > Obsidian vault with backlinks > Wiki starting from index.md > Plain English Q&A over everything It uses two passes. 1. First, it parses code structure without an LLM. 2. Second, Claude subagents extract concepts from docs and images in parallel. Results merge into a single queryable graph. Every connection is tagged as extracted, inferred, or ambiguous. You always know what was found vs guessed. The efficiency gain is 71.5x fewer tokens per query compared to reading raw files. Subsequent queries read the compact graph, not your entire codebase. Open source.

The problem starts with hundreds of saved files. Claude Code can search them, but it searches one by one, slow and blind to how any of them actually connect to each other. The fix is a three-part stack. Graphify maps your files into connected ideas. Obsidian holds that map somewhere you can actually browse it. Claude Code then reads the map instead of digging through the raw pile of files every time. One command bridges the three: graphify --obsidian. Every idea becomes its own note, auto-linked to whatever it relates to, instead of sitting as an isolated file nobody remembers exists. Node size in that map is not decorative. It equals connection count, so the most important ideas surface themselves visually instead of you having to guess which files actually matter. Bare notes still need one more pass. Tell it to pull the source docs in and link every note back to where it came from, so the map stays traceable instead of turning into disconnected fragments. The last step is just moving it where you actually work: move this vault into my main vault, in its own folder. Done in under a minute, and the hundreds of files you started with are now a map instead of a pile. Bookmark this before your next Claude Code session goes back to searching files one by one.