MCP server

Designfit MCP: Check an AI-Built Page Against Its Figma Frame

Designfit compares a rendered front end with its Figma frame using tokens and geometry instead of pixels, then returns a fix list the coding agent can act on.

Run
npm install -g designfit
View on GitHub

Stars, license and last push from GitHub. Install count as reported by the open skills directory.

What is Designfit?

Designfit is an MCP server that validates an AI-built front end against its Figma design. It extracts the frame, measures the rendered page in a browser, and compares design tokens and element geometry rather than pixels. The result is a pass flag, a score and a fix list.

The README opens with a real run on a 360-node Figma frame. The extract tool read the frame from its link, and the validate tool then scored three build iterations: 87, then 89, then 100 pass.

Designfit exists because screenshot diffing makes agents thrash. Anti-aliasing and sub-pixel shifts read as still wrong, so the agent keeps fixing. Designfit measures what a designer would catch instead.

What it measures

  • Design tokens, such as wrong colors and sizes.
  • Geometry, meaning element boxes relative to the screen root, which catches misalignment.
  • Presence, so missing elements are part of the check alongside tokens and geometry.
  • Explicit tolerances, which you can override per call, so the same input always gives the same output.

The fidelity loop

  1. Give the agent a Figma frame link and ask it to implement the design.
  2. designfit_extract returns the design, a component map and the viewport.
  3. The agent builds the page and tags elements with data-designfit-id.
  4. designfit_validate measures the render and returns violations.
  5. The agent fixes and repeats until the result is pass, then strips the tags.

Scope and fit

Best for

  • Anyone letting an agent build from a Figma link who needs an objective stopping rule.
  • Teams already using the Figma MCP to read a design who want a check after the build.
  • Setups that read designs through Framelink and want a conformance step once the page exists.

Not for

  • Responsive work. Version 1 covers one viewport, and multi-breakpoint checks are on the roadmap.
  • Judging taste. A perceptual fallback is only planned, so look to a Figma use skill for guidance on intent.

Use it with UiChemy

Designfit checks code an agent wrote against Figma. UiChemy's Figma to WordPress takes the other path and converts the same design into an editable WordPress page. Teams comparing both can click Get Started on uichemy.com.