Your repo as a knowledge graph.

01Install

Graphify ships as a single Python package + CLI (graphifyy on PyPI). No account, no API keys, nothing leaves your machine. Install the tool, then graphify install registers the /graphify skill with the AI coding assistants it detects: Claude Code, Cursor, GitHub Copilot, Codex, Gemini CLI, Aider, and 17 in total.

$ uv tool install graphifyy✓ installed graphify # alternatives:# pipx install graphifyy# pip install graphifyy

$ graphify install✓ detected Claude Code, Cursor, GitHub Copilot✓ /graphify skill registered · 17 assistants supported

02Build the graph

In your assistant, run /graphify . to map 36 code languages plus Markdown, PDFs, Office docs, SQL schemas, live PostgreSQL, and Terraform into one knowledge graph. Three files land in graphify-out/, and from then on your assistant queries the graph instead of grepping.

# in your assistant (Claude Code, Cursor, Copilot, …):/graphify . ✓ graphify-out/graph.html        interactive graph✓ graphify-out/GRAPH_REPORT.md   architecture report✓ graphify-out/graph.json        machine-readable graph # re-scan only what changed:   /graphify . --update# deeper multi-pass analysis:  /graphify . --mode deep

03Query & traverse

Ask in plain English from your terminal. Answers come back as explicit paths with real file:line citations. Every relation is tagged EXTRACTED, INFERRED, or AMBIGUOUS, so you always know what's grounded in code and what's a guess.

$ graphify query "what connects auth to the database?"AuthService → SessionStore → DatabasePool        [EXTRACTED]  src/auth/service.py:42 → src/db/pool.py:17 $ graphify path "UserService" "DatabasePool"UserService → UserRepository → DatabasePool      (2 hops) $ graphify explain "RateLimiter"RateLimiter · class · src/middleware/rate_limit.pycalled by ApiGateway, WebhookHandler · calls RedisClient

04Use with MCP

The MCP server exposes the graph to any MCP client over stdio, or as a shared HTTP server so several assistants read one graph. It provides 8 tools, including query_graph, get_node, get_neighbors, shortest_path, list_prs, get_pr_impact, and triage_prs.

$ python -m graphify.serve graphify-out/graph.json✓ MCP server ready (stdio) · 8 tools:  query_graph · get_node · get_neighbors · shortest_path  list_prs · get_pr_impact · triage_prs · … # shared HTTP server (one graph, many clients):$ python -m graphify.serve graphify-out/graph.json --http --port 7777

Or register it declaratively. For Claude Code, add this to .mcp.json in your project root:

{  "mcpServers": {    "graphify": {      "command": "python",      "args": ["-m", "graphify.serve", "graphify-out/graph.json"]    }  }}

05Review PRs against the graph

graphify prs maps every open pull request onto the graph: which nodes it touches, where PRs overlap, and which pairs carry merge risk. Use --triage to get a review order and --conflicts to catch collisions before they land, locally or in CI.

$ graphify prs#482  refactor: extract PaymentGateway   touches 12 nodes · 3 shared with #479#479  fix: retry logic in RedisClient    touches 4 nodes $ graphify prs --triage#482  HIGH  overlaps #479 on RedisClient · review first#479  LOW   isolated change $ graphify prs --conflicts#482 ↔ #479  both modify RedisClient.retry · merge risk

Next steps

Deeper reference material lives in the open-source repo.

• GitHub repoSource, issues, and the full README (MIT) · on GitHub
• MCP server referenceAll 8 tools, stdio & shared-HTTP setup · on GitHub
• CLI & Python packagegraphify query / path / explain / prs · on GitHub