Knowledge graphs built on this server from real code with tree-sitter AST parsing — no LLM, no API key, nothing leaves the machine. Click a graph to open the interactive viewer: pan, zoom, search nodes, filter by community, click a node for its neighbours and edges.
1,502 nodes · 3,752 edges · 77 communities · every edge tagged EXTRACTED or INFERRED with a confidence score
Optima (v1 app) — node-level map965 nodes · 2,291 edges · 79 communities
graphify — self-map, aggregated view14,374 nodes · 30,909 edges · 686 communities · rendered as community clusters, because a graph this size is above the 5,000-node node-level render limit
Graphs are also served over the MCP Streamable HTTP transport, so a remote agent (Claude Code, Codex CLI, Hermes, any MCP client) can query the graph without installing graphify locally. API key required — sent as X-API-Key: <key> or Authorization: Bearer <key>.
endpoint : https://graphify.danisha.app/mcp
tools : query_graph · get_node · get_neighbors · shortest_path · list_prs · get_pr_impact · triage_prs
graphify --version # 0.9.66 (venv /srv/danisha/graphify/.venv)
graphify extract <dir> --code-only # local AST extraction, no LLM, no key
graphify cluster-only <dir> # communities + GRAPH_REPORT.md + graph.html
graphify query "how does X work" --graph <dir>/graphify-out/graph.json
graphify explain "NodeName" --graph <dir>/graphify-out/graph.json
graphify path "A" "B" --graph <dir>/graphify-out/graph.json
graphify update <dir> # re-index changed files (free)
Self-hosted Graphify from Graphify-Labs/graphify (Apache-2.0), installed with all extras. Viewer library (vis-network) is vendored locally, so the pages need no external CDN.