Skip to content
Docs / Tools reference

Tools reference

Astron exposes six MCP tools. Every response is short on purpose, usually 600 to 1,200 tokens, so it fits in context next to your code.

get_design_guide and search_references are on every plan. The other four need Pro or Team; on Free they return an error that names the plan, which agents relay to you.

get_design_guide

Returns the guide for a kind of screen or product: type scale, spacing, palette and the rules a designer would hold you to.

When to call: First, before writing any markup for a new screen.

Parameters

kindrequired
string
The screen or product, in plain words: pricing-page, settings, fintech dashboard.
stack
string
Optional hint so rules use your vocabulary: tailwind, css, swiftui.
guide
string
Force a specific guide by slug instead of letting Astron choose.

Example

{
  "method": "tools/call",
  "params": {
    "name": "get_design_guide",
    "arguments": { "kind": "pricing-page", "stack": "tailwind" }
  }
}

search_references

Finds annotated reference screens. Each result has a thumbnail URL and the annotations designers wrote for it.

When to call: After a guide, to see the pattern done well before building it.

Parameters

queryrequired
string
What you are looking for, in plain words.
limit
number
Results to return, 1 to 12. Default 6.
density
"airy" | "balanced" | "dense"
Filter by information density.

Example

{
  "method": "tools/call",
  "params": {
    "name": "search_references",
    "arguments": { "query": "three plan pricing, one recommended", "limit": 2 }
  }
}

load_skillPro

Loads a focused playbook for a moment agents often get wrong, such as empty states or form validation.

When to call: When the task matches a skill name; agents usually pick this from the tool description.

Parameters

namerequired
string
Skill slug, for example empty-states or data-tables. 42 skills on Pro, 10 on Free.

Example

{
  "method": "tools/call",
  "params": { "name": "load_skill", "arguments": { "name": "empty-states" } }
}

critique_uiPro

Reviews markup against the active guide and returns findings by severity, each citing the rule it breaks.

When to call: After a build, on the file that changed. Findings are capped at eight.

Parameters

markuprequired
string
The component or page source. Up to 40 KB.
guide
string
Guide slug to review against. Defaults to the last guide returned in this session.

Example

{
  "method": "tools/call",
  "params": {
    "name": "critique_ui",
    "arguments": { "markup": "<section class=\"hero\">…</section>", "guide": "editorial-saas" }
  }
}

get_palettePro

Returns a neutral ramp and one accent, with contrast ratios for each text pairing.

When to call: When the guide leaves colour open, or for a dark variant.

Parameters

moodrequired
string
A word or two: calm, fintech, warm editorial.
scheme
"light" | "dark"
Default light. Dark palettes use charcoal, not navy.

Example

{
  "method": "tools/call",
  "params": { "name": "get_palette", "arguments": { "mood": "fintech", "scheme": "dark" } }
}

get_type_scalePro

Sizes, line heights and tracking from display to caption, tuned for a typeface.

When to call: When setting up a design system or replacing a default scale.

Parameters

family
string
Typeface name, for example Geist or Inter. Default Geist.
use
"marketing" | "product"
Marketing scales go larger at the top; product scales are tighter.

Example

{
  "method": "tools/call",
  "params": { "name": "get_type_scale", "arguments": { "family": "Geist", "use": "product" } }
}