Rebuild a section from a live site

With the MiroMiro tools connected (Claude, ChatGPT, Cursor, Codex) use extract_code as below. Without them (OpenClaw and other agents) call the REST API with the same parameters as query string:

curl -s "https://miromiro.app/api/v1/code?url=linear.app&section=pricing&framework=tailwind" \
  -H "Authorization: Bearer $MIROMIRO_API_KEY"

No key yet? The user gets one free at https://miromiro.app/api/onboarding (300 credits a month, no card). Ask them for it; never invent one. A call costs 25 credits.

Use the extract_code tool. It resolves the page's real CSS cascade and returns a self-contained component with the site's true colors, fonts, spacing and copy, so you adapt real values instead of guessing from a screenshot.

Picking the section

When the user names a part of the page ("their pricing table", "the footer", "the nav"), pass it as section: one of hero, header, footer, pricing, features, testimonials, faq, cta. MiroMiro finds that block on the page itself, so you never have to see the markup. Never invent a CSS selector like .pricing - a selector that matches nothing falls back to the hero with a note, and the guess still costs a call.

With no section, extract_code auto-detects the hero. Every response also returns a sections array listing the other extractable sections with a human-readable label and a ready-to-use selector: when the part the user wants is not one of the named sections, find it there and call again with that selector.

Choosing the framework

Match the user's project, don't default blindly:

• tailwind (default) — Tailwind utility classes in the markup
• jsx — a React component; pass name to set the component name. Repeated structures (tiers, cards, list items) come back as one template plus a const items = […] array rendered with .map() — use the data array, don't re-expand it.
• vue — a single-file component with <script setup>; name applies here too, and repeated structures collapse into a v-for the same way
• html — plain HTML plus a companion css string
• elementor — an importable Elementor template for WordPress. code is the whole template JSON: have the user save it as a .json file and import it under Templates > Saved Templates > Import.
• shopify — a Shopify theme section. code is the whole Liquid file: save it as sections/<name>.liquid in the theme code editor, then add the section from the theme editor. Not in a Custom Liquid block - that block is capped at 50 KB and cannot hold the section's schema.

Elementor and Shopify output arrives as one styled HTML block, edited as code rather than as native widgets, with responsive and hover styles included and the CSS scoped so it cannot clash with the theme. Hand the file over complete and unmodified; css is null because the file already carries its styles.

For tailwind, jsx and vue, the css field may also carry the section's @keyframes — add those to the project's global stylesheet or the animations silently do nothing.

If the repo you are working in already uses one of these, use it without asking. A user who mentions WordPress, Elementor, Shopify or a theme wants the importable file, not a component.

Light or dark

Pass theme: "dark" when the user is looking at the site in dark mode or asks for the dark version; the response's theme field reports which one you actually got, so tell the user when it differs from what they see. Sites that switch themes purely in JavaScript cannot be selected this way.

When a result comes back thin

The API reads static HTML and CSS by default, and the response tells you how it went: summary.styledCoverage (0-100) is the percentage of elements the CSS cascade resolved styles for, and when the page is client-rendered or styles itself at runtime the response includes a hint field saying what to do. When you see the hint (or coverage is low), call again with render: true — it runs the page in a real browser so its JavaScript executes first. It is slower and costs 5 extra credits, so treat it as the retry, not the default. With render: true you can also pass screenshot: true (or "full") to get a URL of the rendered page, useful to show the user what was extracted.

After you get the code

Adapt rather than paste verbatim:

• Keep the extracted colors, fonts, spacing and copy — that fidelity is the whole point.
• Rewrite the structure to fit the target project's conventions and component patterns.
• Swap any absolute asset URLs for local assets, or keep them if the user wants a reference build.
• The response also carries the design tokens that section actually uses. Wire those into the project's theme instead of hardcoding hex values, when the project has a theme.

Remind the user to check a site's terms before reusing its assets or copy commercially. Extracting for reference and learning is fine; republishing someone's content is theirs to license, not yours to assume.