For the complete documentation index, see llms.txt.
Tools
The seven plagio MCP tools: what each one takes, returns and costs. All of them only read.
Every tool is read-only: it changes nothing on the sites it opens and nothing in your account. Captures spend credits; the three library and reference tools are free.
capture_component
Renders a page and turns one component into a prompt: an HTML and CSS snapshot that renders the same, tokens, structure, hover states, animations and a screenshot. 10 credits, 5 from cache.
| Parameter | Type | Required | Description |
|---|---|---|---|
url | string | Yes | The page the component lives on. |
describe | string | One of the two | The component in plain words, such as "pricing card" or "the Get started button". Quoting visible text is the surest way to hit it. |
selector | string | One of the two | A CSS selector, when you already know it. |
stack | enum | No | react-tailwind (default), react-css, vue, html. |
viewport | enum | No | desktop (1280), laptop (1440), tablet (834), mobile (390). |
fidelity | enum | No | responsive (default) or pixel. |
motion | boolean | No | Include animations. Default true. |
format | enum | No | compact (default): a short prompt plus a bundle with the full snapshot. full: everything inline, often over 100k characters. |
fresh | boolean | No | Skip the cache and render again. |
capture_page
Maps a whole page: its sections with a selector for each, the tokens, a screenshot and a bundle with the complete snapshot. Then capture the sections you need with capture_component. 10 credits.
Takes url, stack, viewport, format and fresh, as above.
extract_brand
Reads one to five pages of a site and returns its design language: weighted colors with roles, hover and focus colors, the type scale with real font names, layout measures, breakpoints, spacing, radii, shadows, the logo and component specimens. It also returns a Tailwind theme, CSS variables, W3C design tokens and a DESIGN.md to save at the root of the project. 10 credits a page.
| Parameter | Type | Required | Description |
|---|---|---|---|
url | string | Yes | The page to start from. The homepage is usually best. |
pages | integer | No | 1 to 5 pages of the same site, following its own links. Default 2. |
paths | string[] | No | Up to 4 specific paths to read after url, such as ["/pricing"]. |
darkMode | boolean | No | Also read the dark theme and return it as a second color set. |
format | enum | No | summary (default) or full, which adds specimens with their HTML and CSS. |
fresh | boolean | No | Read the site again. |
extract_motion
Captures animation and interaction instead of looks: CSS keyframes with timing and easing, JavaScript animations turned into keyframes, elements that move on their own, scroll entrances and draggable elements. 10 credits.
Takes url, and optionally describe or selector to focus on one element, and fresh.
find_references
Suggests hand-made sites to follow from plagio's curated library, with page URLs, screenshots, colors and fonts. Free.
| Parameter | Type | Required | Description |
|---|---|---|---|
page | enum | No | home, pricing, about, customers, blog, changelog, careers, docs, contact, product. Default home. |
industry | enum | No | The user's industry, when it is clear. |
query | string | No | Words about the look or product, such as "dark" or "editorial". |
limit | integer | No | 1 to 12. Default 6. |
list_saved_captures
Lists what you saved with the browser extension: components, pages and recorded flows. Free.
Takes an optional query and kind (component, page or flow).
get_saved_capture
Reads one saved item without opening the site again. For a flow it returns the map of its steps; ask for each one with step. Free.
| Parameter | Type | Required | Description |
|---|---|---|---|
id | string | Yes | The id from list_saved_captures. |
step | integer | No | For a flow: which step to read. |