Experience scan & code sync

The experience graph is most useful when it reflects reality — both the product that's live and the code behind it. Two mechanisms keep it honest: Experience scan to populate the graph, and code references to tie entities to your codebase.

Experience scan

From Create → Experience scan in Experience, point Crux at a live site URL. Crux visits the site and proposes pages and entities from what it finds — routes, navigation, and page regions — which you review before they join the graph.

No live site yet? Give the scan a prompt instead and Crux hypothesises a structure — a plausible set of pages and features for the product you describe. It's a starting point to refine, not a finished IA.

You can also run a scan from the browser extension on the tab you're viewing, after linking a project.

Code references

Coding agents (Cursor, Codex, Windsurf, and others) connect to Crux over MCP — copy the config from Settings → Setup MCP. Once connected, agents register code references as they work: which files implement which entity, tagged in the DOM with the entity's crux-attr short ID.

Each entity then shows a sync status:

Build with MCP

Build with MCP hands an entity to your coding agent: its description, structure, evidence, and short ID travel with it, so the agent builds the right thing and tags it as it goes. When it finishes, the entity flips to Synced.

Next

• Usability scan — score the same page against the usability benchmark
• MCP, overlay, and the browser extension
• The experience graph