graphify: Turn Any Codebase Into a Queryable Knowledge Graph for Claude Code

Compound Engineering: Orchestrate Claude Code, Codex • Orca: The ADE for Running Claude Code, Codex, and Cursor in Parallel

What Is graphify? #

graphify answers a question every AI coding agent eventually runs into: once a codebase gets big, grepping and re-reading files on every prompt gets slow and lossy. graphify’s fix is to build a knowledge graph of the whole project once — code, docs, SQL schemas, configs, PDFs, images, even video — and let the agent query it instead of re-reading everything.

🔗 GitHub: https://github.com/Graphify-Labs/graphify 🌐 Homepage: https://www.graphify.com

Built by Graphify Labs (Y Combinator S26) and first pushed in April 2026, graphify had already reached 97,700+ GitHub stars by late July 2026 — one of the fastest-growing dev tools of the year, and notably undercounted on GitHub’s own trending page (which showed only ~8,590 at one point the project’s own topic research caught, versus the real API-verified count).

Three things it’s careful to state up front:

• Code parsing is fully local — tree-sitter AST parsing, deterministic, no LLM call, nothing leaves your machine. Only the optional semantic pass over docs/PDFs/media calls a backend.
• Every edge is labeled EXTRACTED (explicit in the source) or INFERRED (resolved by graphify), so you know what’s certain versus derived.
• It’s not a vector index — no embeddings, no vector store. A real graph you can traverse, trace paths in, and ask “what connects to what.”

Get Started (30 Seconds, Per the Project) #

uv tool install graphifyy      # install the CLI (or: pipx install graphifyy)
graphify install               # register the skill with your AI assistant

Then, inside your AI assistant:

/graphify .

That produces three outputs:

graphify-out/
├── graph.html       open in any browser — click nodes, filter, search
├── GRAPH_REPORT.md  key concepts, surprising connections, suggested questions
└── graph.json       the full graph — query it anytime without re-reading your files

Works in Claude Code, Cursor, Codex, Gemini CLI, GitHub Copilot, and 15+ other assistants.

Naming note: the PyPI package is graphifyy (double-y) — the CLI command itself is still graphify. The project explicitly flags that other graphify*-named PyPI packages aren’t affiliated.

Querying the Graph #

Once built, you ask questions instead of reading files:

graphify explain "APIRouter"
graphify path "FastAPI" "ModelField"
graphify query "what connects auth to the database?"

The project’s own example run against the FastAPI codebase returns a real node with its source location, community cluster, connection count, and each connection tagged EXTRACTED or INFERRED — and a 3-hop path between two specific classes, traced edge by edge.

What It Does #

Common Commands #

/graphify .                        # build graph for current folder
/graphify ./docs --update          # re-extract only changed files
/graphify . --cluster-only         # rerun clustering without re-extracting
/graphify . --wiki                 # build a markdown wiki from the graph

/graphify query "what connects auth to the database?"
/graphify path "UserService" "DatabasePool"
/graphify explain "RateLimiter"

/graphify add https://arxiv.org/abs/1706.03762   # fetch a paper and add it to the graph

graphify hook install              # auto-rebuild the graph on every git commit
graphify prs --triage              # AI ranks your PR review queue against the graph
graphify prs --conflicts           # flags PRs sharing graph communities — merge-order risk

Benchmarks (Per the Project’s Own Published Results) #

The project states every system ran on the same harness with the same model and budget, scored by an LLM judge cross-validated against a second judge (90.6% agreement, Cohen’s kappa 0.81). Full methodology and reproduction commands are in the project’s own BENCHMARKS.md — these numbers are self-reported, not independently reproduced here.

Install #

Requires Python 3.10+.

# macOS (Homebrew)
brew install python@3.12 uv

# Ubuntu/Debian
sudo apt install python3.12 python3-pip pipx

# Windows
winget install astral-sh.uv

Then install the package itself:

uv tool install graphifyy
graphify install

Ignoring Files #

graphify respects .gitignore automatically, and merges in an optional .graphifyignore (same syntax, ! negation supported) that always excludes more, never re-includes something .gitignore already excluded:

# .graphifyignore
node_modules/
dist/
*.generated.py

# only index src/, ignore everything else
*
!src/
!src/**

Use Cases #

1. Orienting an AI Agent in an Unfamiliar Codebase #

Run /graphify . once on a large repo, and let the agent query graph.json for structure instead of repeatedly re-reading files across a session.

2. Tracing How Two Systems Connect #

graphify path "ModuleA" "ModuleB" gives an explicit hop-by-hop chain instead of manually chasing imports across files.

3. Auditing PR Merge-Order Risk #

graphify prs --conflicts flags open PRs that touch the same graph communities, surfacing likely merge conflicts before they happen.

4. Building Docs From the Codebase Itself #

/graphify . --wiki generates a markdown wiki directly from the extracted graph rather than hand-written documentation drifting out of sync.

Related Repositories #

Related Articles #

• Compound Engineering: Orchestrate Claude Code, Codex — another skill-based approach to structuring AI coding workflows
• Orca: The ADE for Running Claude Code, Codex, and Cursor in Parallel — a different layer of AI-coding infrastructure (orchestration vs. codebase understanding)

Conclusion #

graphify treats “does the AI actually understand this codebase” as a graph problem rather than a search problem — local, deterministic tree-sitter parsing feeding a real traversable graph, with every edge labeled by how certain it is. Nearly 100,000 stars within four months, and a GitHub trending page that undercounted it by roughly 10x at one point, both point to real, fast-growing demand for this specific approach over plain RAG/vector search for code.

Best for: Developers working in large or unfamiliar codebases with an AI coding assistant, who want the assistant to query a structured map of the project instead of repeatedly re-reading files.

GitHub: https://github.com/Graphify-Labs/graphify

Recommended Infrastructure for Self-Hosting #

If your team wants to serve one shared graph over HTTP instead of running graphify locally per-developer:

• DigitalOcean — $200 free credit for 60 days across 14+ global regions, a straightforward host for a team-shared graph server.
• HTStack — Hong Kong VPS with low-latency access from mainland China. This is the same IDC that hosts dibi8.com — battle-tested in production.

Affiliate links — they don’t cost you extra and they help keep dibi8.com running.

Last updated: 2026-07-29

References & Sources #

• graphify
• graphify homepage
• graphify benchmarks
• Claude Code
• tree-sitter