Claude in Chrome
Ask Claude about a part while you look at it in the modeler or on its catalog page.
Claude in Chrome is a browser extension. Its agent works on the page you have open: it reads the page, takes screenshots, and clicks and types as you would.
Open its side panel on the modeler, or on a device's catalog page, and ask, e.g. "the height doesn't make sense, fix it". It can help in two ways.
On the page
The Agent brief card lists everything an agent needs, in millimetres:
- a summary of the device;
- every measurement, with its source;
- each face's size;
- the ports and their positions;
- your open annotations;
- how to connect the MCP server.
The face editor states how many millimetres one pixel is, so positions read off a screenshot come out right. Every form field has a clear name, so the agent can change values the way you would.
Copy context for your agent copies the brief as Markdown, to paste into any chat.
Over MCP
If you've added Rack Maker as a connector in Claude, it works in Claude in Chrome too. Your agent's edits then apply directly to your draft and show up live. It works through your annotations and measurement form, resolving each one with a note. See Using AI agents for how to connect.
Datasheet uploads
How your agent keeps a datasheet PDF with a device - from a public URL, or from your computer by a one-time upload link - and what each answer means.
Walkthrough: a DGX Spark
A worked example - model an NVIDIA DGX Spark from your photos with Claude Code, check it face by face, and publish it.