Graphify Labs, a San Francisco startup fresh out of Y Combinator's Summer 2026 cohort, has released an open-source tool that transforms entire codebases into traversable knowledge graphs, accumulating 122,800 GitHub stars as of September 30 and 7.67 million downloads as of September 27 since its July launch. The numbers suggest rapid adoption, though the speed has prompted some online skepticism about whether the momentum is entirely organic.
The company claims engineers at Shopify, Datadog, J.P. Morgan, American Express, and Vanguard now use the tool, though these claims are based on company marketing materials. Those names appear as logos on Graphify's marketing site without independent confirmation. Public adopters verified through linked posts include Rootly AI Labs, the YC-backed agent platform Superagent, and Hong Kong University of Science and Technology's KnowComp research lab.
At its core, Graphify is a command-line utility that runs on-premises. Engineers point it at a directory containing code, documentation, PDFs, SQL schemas, or infrastructure configs, and the tool parses everything locally using tree-sitter across 36 programming languages. What emerges is a typed knowledge graph with labeled nodes and edges, packaged as an interactive HTML visualization, a markdown summary, and raw JSON.
The appeal for AI coding assistants is straightforward. Rather than dumping entire files into a prompt window and burning through context windows, developers can feed these structured graphs to tools like Claude Code, Cursor, or GitHub Copilot. Graphify integrates via the Model Context Protocol, either through a hosted OAuth endpoint or a local server installed with a single command. The company cites developer walkthroughs claiming "71.5× fewer tokens" when pairing Claude Code with Graphify's graph context versus raw file ingestion on a nearly half-million-token codebase.
Each connection in the graph carries a provenance label. EXTRACTED means it came from the abstract syntax tree. INFERRED signals a model generated the link. AMBIGUOUS flags uncertainty. The tool exports to Neo4j, FalkorDB, GraphML, and Obsidian. "No code or data ever leaves your environment," the company states on its Y Combinator profile.
Safi Shamsi, who holds a master's in data science from the University of Birmingham with a thesis on knowledge-graph-powered retrieval, founded Graphify and released the Apache 2.0-licensed engine on July 1, 2026. The startup joined YC's Summer 2026 batch under partner Jared Friedman.
According to Dealroom, Graphify hit 100,000 GitHub stars on August 1, becoming the first YC company to cross that threshold during its own batch. The repository now holds 122,800 stars and 11,800 forks as of late September. PyPI data shows 7.67 million total downloads of the graphifyy package through September 27, with 1.65 million in the prior month alone.
That velocity has raised eyebrows. Online discussions in June and August questioned whether the star accumulation reflected genuine adoption or paid promotion, though no reporting has substantiated those suspicions. One August post surfaced security concerns about installer scripts fetching components from URLs without pinned checksums. The company has not publicly addressed the allegations.
Beyond the free open-source CLI, Graphify sells a hosted enterprise tier with pull-request review and formal verification capabilities. The verification module compares old and new code behavior differentially, classifying each change as Proven, Changed, or Unknown, then writes a failing-input test when it detects a behavioral shift. Pricing starts at free for the core tool, $10 monthly for Pro, and $20 per seat for teams up to 100 users, then $28. Enterprise pricing is custom. The company offers self-hosted deployment and has begun a SOC 2 Type II engagement. Open-source projects under MIT or Apache licenses get free access.
Y Combinator lists Graphify's team size as one, though LinkedIn shows "2–10 employees." The company declined to clarify headcount beyond the YC listing.
Graphify enters a nascent but growing field. Sourcegraph's Cody assistant uses an internal "Code Graph" for code relationships and context, according to its documentation. Microsoft Research's GraphRAG project builds LLM-derived knowledge graphs for retrieval-augmented generation, emphasizing graph-based reasoning over pure vector search. GitNexus, another open-source project, also positions itself as an MCP-native code knowledge graph for AI agents, with community discussions surfacing earlier this year.
"Migrating a large codebase with AI is like changing a train's engine while it is moving," Shamsi wrote in the company's Y Combinator launch post in late August. "Graphify lets you do it without stopping the train."
In an August LinkedIn post, Shamsi said the startup had reached "4,000,000+ downloads" and "1,500+ customers," adding, "The second number is the one I'm proud of." The company has not announced funding beyond Y Combinator's standard investment. Crunchbase lists a pre-seed round with YC as lead but restricts details behind a paywall.
Whether the early traction translates into sustainable enterprise revenue remains an open question. For now, Graphify's growth trajectory reflects a broader shift in how developers expect AI assistants to understand their code—not as isolated files, but as interconnected systems with traceable relationships.
