For engineers using coding agents
Understand what your coding agent built
An agent can change services, endpoints and tables across dozens of files in one session, and its summary reads like a changelog. Connect it to ArchDocs over MCP and ask it to draw what it changed. You get one diagram of those pieces and how they connect, to check against the diff.
Sound familiar?
- The agent says it's done. I can't tell which services it touched.
- The diff is huge and I don't know where to start reading it.
- It added a Kafka consumer somewhere. I only found out from the config change.
Every piece and how it connects
REST and gRPC endpoints become cards with their parameters, responses and errors. Services and queues are references to objects in your catalog. Tables show their columns and keys.
The objects your team already documents
The agent can list your catalog before it draws, so the Ledger DB in its diagram is the same object your other docs use. Open that object and its Used in docs tab shows every document that references it.
Correct it where it's wrong
The agent writes the diagram as YAML. Fix a label on the canvas, edit the YAML panel next to it, or ask the agent to patch the document.
Reviewing a feature an agent built
Connect the ArchDocs MCP server to Claude Code or Cursor.
When the agent finishes, ask it to diagram what it built.
Read the diagram first and note anything you didn't expect.
Review the diff with that diagram open beside it.