MCP App Store
by KeenEthics
Uiprobe icon

Uiprobe MCP Server for ChatGPT

By Uiprobe

Overview

Uiprobe automatically catches visual differences between your Figma designs and live web pages: layout and spacing, typography, color, size, and content. Every finding shows the expected value and the actual value. Give ChatGPT a Figma frame URL and a page URL, and it runs the comparison and goes through the findings with you. In Codex, it can find the element in your code, fix it and run the probe again to see what's still off. Sign in with Figma at app.uiprobe.io first. The page has to be reachable from the internet.

How to connect

  1. ChatGPT: Open the Uiprobe listing in the ChatGPT app directory and choose Connect. ChatGPT listing Uiprobe exposes 8 tools here.

Tools

dismiss_findings

ChatGPT
Dismiss findings of a completed probe run by id (from get_probe), up to 50 per call, e.g. differences the user says are intentional. Dismissed findings are hidden for every viewer of the probe and stay dismissed on its future runs. A dismissal can always be undone with restore_findings. Ask the user before dismissing. probeVersionId is the run the ids came from. A grouped finding's id dismisses the whole group; wholeGroup true does the same for a child id. Findings already dismissed are left as they are.

get_finding_images

ChatGPT
Return images of findings of a completed probe: for each finding id (from get_probe), a crop of the Figma design and a crop of the captured webpage around the finding's elements, with them outlined. Up to 5 findings per call. Use it when you need to see a visual difference, not for every finding. side limits the images to the design or the page.

get_probe

ChatGPT
Fetch the findings of an existing probe by id (any probe the calling user can access, including team probes), once get_probe_status reports it COMPLETED. Returns what the probe viewer shows: findings hidden by the probe's finding filters and dismissed findings are left out and counted in hiddenCount and dismissedCount; includeHidden and includeDismissed add them after the others, flagged. Findings are paged: pass the previous result's nextOffset as offset to get the next page. The result names the Figma frame (figmaUrl) and webpage (webpageUrl) the probe compares.

get_probe_status

ChatGPT
Return only the status of a probe's latest run (PENDING, RUNNING, COMPLETED or FAILED), without findings. With waitForResult true it waits on the server up to waitForResultSeconds (default 30, max 60) for a PENDING or RUNNING probe to finish. Repeat the call, each time with waitForResult true, until the status is COMPLETED or FAILED, then call get_probe for findings. If the probe is still PENDING after about 5 minutes of polling, stop and tell the user, with the viewer URL.

list_probes

ChatGPT
List the probes the calling user can access, personal and team; team probes carry their teamId. Each entry names the Figma frame (figmaUrl) and webpage (webpageUrl) of its latest run, so a page can be matched to its probe. Optionally filter by a case-insensitive substring of the title and sort by name or creation date.

rerun_probe

ChatGPT
Run an existing probe again against its webpage and return findings. Creates a new version of the probe (its history is kept) using the same Figma design; optionally pass webpageUrl to capture a different URL, e.g. a new tunnel URL. Polls until it finishes unless waitForResult is false. Each call adds a version and uses one run of the user's allowance, so calling it again repeats both. If it returns PENDING or RUNNING, call get_probe_status with waitForResult true until the run finishes, then get_probe for findings.

restore_findings

ChatGPT
Restore dismissed findings of a probe by id, up to 50 per call, so every viewer sees them again, on this run and future ones. Get the ids from get_probe with includeDismissed true, and pass that result's probeVersionId. A grouped finding's id restores the whole group; wholeGroup true does the same for a child id. Findings that are not dismissed are left as they are.

run_probe

ChatGPT
Compare a Figma frame against a live webpage and return findings. Creates a probe, optionally polls until it finishes, and returns status + findings. Each call creates a new probe and uses one run of the user's allowance, so calling it again repeats both; to check a design and page that already have a probe, use rerun_probe. If it returns PENDING or RUNNING, call get_probe_status with waitForResult true until the run finishes, then get_probe for findings. webpageUrl must be publicly reachable; for a localhost page, expose it through a cloudflared quick tunnel and pass the tunnel URL (see the server instructions).

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/uiprobe"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/uiprobe)