HTTP API from CI
Create a key and call the endpoints to start a check on a public URL and read the result. Designed to run on each preview deploy.
Day 4
Add a UI check step to your Claude Code or Cursor workflow
Xpensiv SI is built to work as an MCP server, so a coding agent such as Claude Code or Cursor can call a UI check on your AI-built app, read the findings, and propose fixes inside your existing workflow. The findings come from rule-based checks, and you approve each change before it is applied.
Read the API and MCP docsThree ways to call it
MCP in your editor
Connect Cursor or Claude Code and let the tool request a check and read findings in the same session. Designed as a read-only interface.
Into a pull request
Turn findings into an edit in your own tooling, open a pull request, and a person approves before it merges.
A check inside the loop
- 01
Trigger
CI or your editor starts a check on the preview URL through the API or MCP.
- 02
Read findings
Your tool pulls the fix list, each item written as an instruction with a screenshot.
- 03
Apply in a branch
You or your AI tool make the edits, so the diff is yours to see.
- 04
Review and merge
Open a pull request, a person approves, then re-run the check to confirm the findings are gone.
What is here today
The API and MCP are the intended interfaces for running checks from your workflow. A vision AI model reads the browser measurements and screenshots and returns a fix list; it reports and never edits your code. The page describes how to connect and what the check measures, not a promised result.
Questions answered
Is there an MCP server for UI design fixes?
The API and MCP are the intended interfaces for running checks from your workflow. Once connected, your AI coding tool can request a check and read the findings in the same session; it is designed as a read-only interface.
How do I add a design review step to a Claude Code or Cursor workflow?
Connect Cursor or Claude Code over MCP so the tool can pull findings itself, or call the HTTP API from CI. You then feed the findings into a pull request that a person approves before it merges.
What tool improves the UI of an AI-built app inside my coding agent?
Xpensiv lets your coding tool request a check and read the fix list in the same session. The edits are made in your own tooling and approved by a person; it does not rewrite your app.
Wire a check into your pipeline.
Public pages only. The check reads the live page and does not change your app.