graphify

name	graphify
description	any input (code, docs, papers, images) → knowledge graph → clustered communities → HTML + JSON + audit report
trigger	/graphify

/graphify

Turn any folder of files into a navigable knowledge graph with community detection, an honest audit trail, and multiple outputs: interactive HTML, GraphRAG-ready JSON, and a plain-language GRAPH_REPORT.md.

Usage

/graphify                                             # full pipeline on current directory
/graphify <path>                                      # full pipeline on specific path
/graphify <path> --code-only                          # code files only, no LLM needed
/graphify <path> --no-llm                             # skip semantic extraction (AST only)
/graphify <path> --update                             # incremental - re-extract only new/changed files
/graphify <path> --format json,html,report            # select specific export formats
/graphify <path> --format graphml                     # export graph.graphml (Gephi, yEd)
/graphify <path> --format cypher                      # generate graphify-out/cypher.txt for Neo4j
/graphify <path> --format svg                         # export graph.svg
/graphify <path> --format wiki                        # build agent-crawlable wiki
/graphify <path> --format obsidian                    # write Obsidian vault
/graphify query "<question>"                          # BFS traversal - broad context
/graphify query "<question>" --dfs                    # DFS - trace a specific path
/graphify query "<question>" --budget 1500            # cap answer at N tokens
/graphify add <url>                                   # fetch URL, save to ./raw, update graph
/graphify watch <path>                                # watch folder, auto-rebuild on code changes
/graphify serve                                       # start MCP stdio server for agent access

What graphify is for

graphify is built around Andrej Karpathy's /raw folder workflow: drop anything into a folder - papers, tweets, screenshots, code, notes - and get a structured knowledge graph that shows you what you didn't know was connected.

Three things it does that Claude alone cannot:

1. Persistent graph - relationships are stored in graphify-out/graph.json and survive across sessions. Ask questions weeks later without re-reading everything.
2. Honest audit trail - every edge is tagged EXTRACTED, INFERRED, or AMBIGUOUS. You know what was found vs invented.
3. Cross-document surprise - community detection finds connections between concepts in different files that you would never think to ask about directly.

Use it for:

• A codebase you're new to (understand architecture before touching anything)
• A reading list (papers + tweets + notes → one navigable graph)
• A research corpus (citation graph + concept graph in one)
• Your personal /raw folder (drop everything in, let it grow, query it)

What You Must Do When Invoked

If no path was given, use . (current directory). Do not ask the user for a path.

Follow these steps in order. Do not skip steps.

Step 1 - Ensure graphify-rs is installed

if ! command -v graphify-rs >/dev/null 2>&1; then
  echo "graphify-rs not found. Install with: cargo install graphify-rs"
  exit 1
fi
graphify-rs --version

If the binary is found, print nothing extra and move straight to Step 2.

Step 2 - Build the knowledge graph

Run the full pipeline. graphify-rs handles detection, extraction, building, clustering, analysis, and export in a single command:

graphify-rs build --path INPUT_PATH --output graphify-out

Replace INPUT_PATH with the actual path the user provided.

Available flags:

• --no-llm: skip Claude API semantic extraction (AST-only, free, fast)
• --code-only: only process code files
• --update: incremental rebuild, only re-extract changed files
• --format json,html,report,wiki,svg,graphml,cypher,obsidian: select export formats (default: all)
• --jobs N: control parallelism
• --max-viz-nodes N: maximum nodes in HTML visualization (default: 2000, increase for larger projects)

The command outputs progress with a progress bar and colored status messages.

If the user specified --code-only or --no-llm, pass those flags through.

Step 3 - Present results

After the build completes, read and present key sections from the report:

cat graphify-out/GRAPH_REPORT.md

Present these sections directly in chat:

• God Nodes (highest-connectivity nodes)
• Surprising Connections (cross-community bridges)
• Suggested Questions

Do NOT paste the full report - just those three sections. Keep it concise.

Step 4 - Offer to explore

Pick the single most interesting suggested question from the report and ask:

"The most interesting question this graph can answer: [question]. Want me to trace it?"

If the user says yes, run:

graphify-rs query "QUESTION" --graph graphify-out/graph.json

Walk them through the answer using the graph structure. Each answer should end with a natural follow-up so the session feels like navigation.

---

Keeping the graph current after code changes

This is critical for agentic workflows. When you (or the user) modify code files during a session, the knowledge graph becomes stale. You MUST rebuild it to keep answers accurate.

Rule: After modifying code, rebuild the graph

After you finish a batch of code changes (new files, edited functions, refactored modules), run:

graphify-rs build --path . --output graphify-out --no-llm --update

• --update: only re-extract changed files (fast, uses SHA256 cache)
• --no-llm: skip Claude API (AST-only rebuild is free and fast, ~2-5s)
• This updates graph.json, GRAPH_REPORT.md, and all exports

When to rebuild

• After writing/editing code: rebuild before answering architecture questions about the changed code
• After refactoring: rebuild so community structure reflects the new module boundaries
• After adding new files: rebuild to include them in the graph
• NOT after every single edit: batch changes, then rebuild once. Don't rebuild after a one-line typo fix unless the user asks architecture questions about it.

Automated alternatives

Instead of manual rebuilds, the user can set up always-on monitoring:

1. Watch mode (background process):

   graphify-rs watch --path . --output graphify-out
   

   Auto-rebuilds on file changes with 3s debounce. Best for long coding sessions.

2. Git hooks (per-commit):

   graphify-rs hook install
   

   Rebuilds after every git commit. No background process needed.

3. Claude Code integration (always-on):

   graphify-rs claude install
   

   Writes a PreToolUse hook to .claude/settings.json that reminds you to check the graph before searching files, plus a CLAUDE.md rule to rebuild after code changes.

---

For /graphify query

graphify-rs query "QUESTION" --graph graphify-out/graph.json

Add --dfs for depth-first traversal (trace specific paths). Add --budget N to control output size (default 2000 tokens).

After answering, save the result for the feedback loop:

graphify-rs save-result --question "QUESTION" --answer "ANSWER" --nodes NODE1 NODE2

---

For /graphify add

Fetch a URL and add it to the corpus:

graphify-rs ingest URL --output graphify-out

Then rebuild incrementally:

graphify-rs build --path . --output graphify-out --update

---

For --watch

Start a background watcher that monitors a folder and auto-updates the graph:

graphify-rs watch --path INPUT_PATH --output graphify-out

Code changes trigger AST re-extraction + rebuild automatically (no LLM needed). Press Ctrl+C to stop.

---

For MCP server

Start a stdio MCP server exposing 15 query tools:

graphify-rs serve --graph graphify-out/graph.json

Tools: query_graph, get_node, get_neighbors, get_community, god_nodes, graph_stats, shortest_path, find_all_paths, weighted_path, community_bridges, graph_diff, pagerank, detect_cycles, smart_summary, find_similar.

To configure in Claude Desktop, add to claude_desktop_config.json:

{
  "mcpServers": {
    "graphify": {
      "command": "graphify-rs",
      "args": ["serve", "--graph", "/absolute/path/to/graphify-out/graph.json"]
    }
  }
}

---

For git commit hook

graphify-rs hook install     # install post-commit/post-checkout hooks
graphify-rs hook uninstall   # remove hooks
graphify-rs hook status      # check if hooks are installed

After every git commit, the hook auto-rebuilds the graph (code-only, no LLM).

---

For Claude Code integration

Run once per project to make graphify always-on:

graphify-rs claude install     # write ## graphify section to CLAUDE.md + PreToolUse hook
graphify-rs claude uninstall   # remove the section

---

Additional commands

graphify-rs stats graphify-out/graph.json              # show graph statistics
graphify-rs diff old-graph.json new-graph.json         # compare two graph snapshots
graphify-rs benchmark graphify-out/graph.json          # token efficiency benchmark
graphify-rs init                                       # create graphify.toml config file
graphify-rs completions bash                           # generate shell completions (bash/zsh/fish)

---

Honesty Rules

• Never invent an edge. If unsure, use AMBIGUOUS.
• Never skip the corpus check warning.
• Always show token cost in the report.
• Never hide cohesion scores behind symbols - show the raw number.
• Never run HTML viz on a graph with more than 5,000 nodes without warning the user.