When AI agents need to understand a codebase, they ask us — instead of re-cloning and re-indexing every time.

Public registry of pre-computed code-knowledge graphs. Schema-validated, drift-tracked, MIT/Apache code, $0 to use.

Built on the open Code-Knowledge-Graph Protocol v1.

  • — indexed repos
  • — graph formats
  • 4 packages live
  • 10 upstream integrations in flight

AI agents

One URL = every public code graph we track.

curl https://looptech-ai.github.io/understand-quickly/registry.json

Tool makers

Add a 5-line GitHub workflow. Your users get indexed automatically.

looptech-ai/uq-publish-action ↗

Developers

Register your repo via the wizard or CLI. One PR. Schema validation built-in.

npx @looptech-ai/understand-quickly-cli add

Drop this into your MCP config

Get list_repos, get_graph, search_concepts, find_graph_for_repo as agent tools in Claude Desktop, Codex, Cursor — anything that speaks MCP. 

{
  "mcpServers": {
    "understand-quickly": {
      "command": "npx",
      "args": ["-y", "@looptech-ai/understand-quickly-mcp"]
    }
  }
}

Listed on the MCP Registry as io.github.looptech-ai/understand-quickly.

What is a code-knowledge graph?

A code-knowledge graph is a structured map of a codebase — files, functions, classes, and their relationships — designed for AI agents to query. Tools like understand-anything, GitNexus, and code-review-graph produce them; we aggregate them.

Agent endpoints
GET ./registry.json — canonical index
GET ./schemas/ — JSON Schemas
+ Add your repo Read the docs