Observed — Homepage uses technical jargon without explanation: 'MCP' (Model Context Protocol), 'corpus', 'audit', 'checklist' are not defined for unfamiliar visitors. Example: 'serves a curated UX corpus to coding agents over MCP' assumes knowledge.
See Rubric in 60 seconds.
A real audit, a predicted heatmap, a simulated visitor and live analytics, on our own site.
Mark up a website the way a scribe marked a manuscript.
The rubric was the passage a scribe wrote in red so a reader could follow the text. Rubric does that to a page: before launch, and after it.
UX audit
Hundreds of checks, up to 7 pages. Findings by severity, each with the fix.
See how it works →
Predicted
Heatmap
Where the eye goes, on desktop and mobile, before launch.
See how it works →
Simulated
Probe
BetaDescribe a visitor. A simulated one tries the page and says where it hesitates.
See how it works →
Analytics
Real visitors, no cookies, and a weekly read of what to fix.
See how it works →Four parts. Ready for the client report.
Each finding has four parts: the issue, why it matters, what to do, and the references behind it. Paste the issue into a client report; hand the actions to whoever builds the fix.
- Issue
- Impact
- Actions
- References
The page lacks a low-commitment secondary path — only a single primary action with no fallback for visitors who aren't ready to commit yet.
Visitors not ready to act at the primary level have no way to engage; they leave instead of taking any step at all.
- Add a secondary, lower-commitment option beside or beneath the primary CTA — e.g. "Download the brochure", "Watch a short overview", "Request a callback", or "Read client stories" — matched to what the page actually offers.
- Style it as a text link or outlined button so it doesn't compete visually with the primary action.
- Track completion per path so you understand which secondary options actually convert.
Give it a URL. Get back what to fix first.
Map the pages
A discovery pass finds the pages worth auditing. It's part of the audit.
Walk them like a first-time visitor
Behind a login too: you sign in yourself, nothing typed is stored.
Report what fails
Findings grouped by severity, each with the fix. Watch it happen live, step by step.
The same checks, while it designs.
Claude Code and other MCP clients can call the same checks while they build. One command, approve in the browser, done.
claude mcp add --transport http rubric https://rubric.design/mcp
/plugin marketplace add things-studio/rubric-plugins
/plugin install rubric-audit@rubric
Plans with a clear monthly allowance.
Audits, heatmaps and Probe, done for you on any website, each with its own monthly allowance. Audits your own agent does over MCP count as calls. Annual billing: two months free. Every plan has a 14-day no-questions refund. Prices exclude applicable taxes, calculated at checkout.
Free
For using the rules from your own coding agent.
- 25 MCP calls a day
- The full checklist
- Saved audits: issue, severity
- MCP only, no managed work
Basic
For a freelancer looking after a few websites.
- 5 managed audits a month
- 20 heatmaps · 4 Probes
- 1 seat · extra $2/month
- Analytics on 1 site
- 5,000 pageviews a month
- 2,000 MCP calls a month
- Impact, actions, references
Pro
most includedFor an agency team looking after many websites.
- 25 managed audits a month
- 100 heatmaps · 20 Probes
- 3 seats · extra $2/month
- Analytics on 5 sites
- 50,000 pageviews a month
- 20,000 MCP calls a month
- Impact, actions, references
Enterprise
For a team with a procurement process.
- Custom quotas
- Extra seats
- Invoiced billing
- EU-hosted analytics
Need more this month?
On top of a paid plan. Unused pack allowance ends with your billing month.