Model Context Protocol

Your styles, callable from any AI agent

The StyleRef MCP server lets Claude, Cursor, VS Code, and ChatGPT search the gallery, fetch a compiled style, wrap a prompt, or write a STYLE.md — one tool call, no setup, no login for read tools.

https://styleref.io/api/mcp

Install Read the docs

Four read tools, zero friction

Every tool is read-only and works anonymously. Each response carries the canonical StyleRef URL so styles you use travel with attribution.

search_styles

Search the public gallery by aesthetic, mood, or discipline. Returns compact cards (name, slug, @handle, essence, canonical URL).

get_style

Fetch a style compiled into any format — default prose, FLUX, Midjourney, diffusion, STYLE.md, or JSON. Supports sections + compact for token discipline.

apply_style_to_prompt

Wrap a raw request in a style and get one ready-to-paste prompt for ChatGPT, Claude, Gemini, Midjourney, FLUX, DALL·E, or Stable Diffusion.

get_style_md

Return a complete STYLE.md body for a coding agent to write into a repo root and reference from AGENTS.md.

OAuth 2.1

Connect your account, act on your library

Add StyleRef as a connector and sign in once — your agent can then read your library, save and remix gallery styles, author new styles, and run AI extractions with your plan's credits. Permissions are scoped and shown on the consent screen.

list_my_stylesstyles:read

Your own style library — every style listed with a ref you can pass straight back to get_style.

save_style / fork_stylestyles:write

Save (like) a gallery style, or remix it into your library as an editable private copy with lineage.

create_style / update_stylestyles:write

Author styles from schema-validated blocks, or duplicate an existing one (create_style with from_ref). Edit rename + blocks.

extract_styleextract

AI extraction from text, a named reference (e.g. "Bauhaus"), an image URL, and/or a document (PDF) — any combination. Async; uses extraction credits.

publish_style / set_style_visibilitystyles:write

Publish to the gallery or activate a shareable link through moderation; revoke or remove-from-gallery. Completes extract → refine → publish inside the agent.

validate_style_md

Lint a STYLE.md document against the spec — frontmatter, section ids, hex colors. Works without an account.

Connect in 10 seconds

Pick your client and paste. The endpoint is the same everywhere.

claude mcp add --transport http styleref https://styleref.io/api/mcp

One call, styled output

No copy-pasting style blobs. The agent composes the prompt for you.

You:  Make me a Midjourney prompt for a product shot in the "80s-Inspired Poster" style.

Claude:  [calls apply_style_to_prompt · ref="1an6edgp-c42c0cba5315" · target="midjourney"]

  product shot of a ceramic mug on a marble surface,
  retro print advertising, oversized high-contrast serif headline,
  sepia-cream base with a bold gold accent, filmic retro print-ad
  lighting, film grain and slight vignetting --ar 4:5 --v 6

  — Usage (midjourney): Subject leads; style keywords and --parameters trail.
  Generated by StyleRef.io — https://styleref.io/share/1an6edgp-c42c0cba5315?utm_source=mcp

Resources

URI-addressable, read-only data for hosts that support MCP resources.

styleref://gallery/{slug}Compiled spec for a gallery style

styleref://gallery/{slug}/style.mdSTYLE.md body for a gallery style

styleref://schemaThe StyleRef style schema (authoring guide)

Prompts

Reusable interaction templates that standardize how agents use StyleRef.

use-styleGenerate something strictly following a chosen style

pick-a-styleGuided discovery: discipline + mood → 3 candidates

setup-style-mdFetch a STYLE.md, write it, and wire it into AGENTS.md

No MCP? No problem

The same styles over plain HTTP

Every read tool is also a public REST endpoint — anonymous, rate-limited, and described by an OpenAPI 3.1 spec at /api/v1/openapi.json. Use it from custom GPTs, Zapier, or the npx styleref CLI — and bring your own token (npx styleref login) to pull your private styles.

GET/api/v1/styles?query={q}

Search the gallery (JSON cards)

GET/api/v1/styles/{ref}?format={fmt}

Compiled spec — default, flux, midjourney, diffusion, stylemd, json

GET/api/v1/styles/{ref}/style.md

Complete STYLE.md body

POST/api/v1/style-md/validate

Lint a STYLE.md document

# Search, then pull a style compiled for FLUX
curl "https://styleref.io/api/v1/styles?query=warm+editorial&limit=3"
curl "https://styleref.io/api/v1/styles/1an6edgp-c42c0cba5315?format=flux&compact=1"

# Or from the terminal — writes STYLE.md + an AGENTS.md note
npx styleref init 1an6edgp-c42c0cba5315

Browse the gallery, then call it

Find a style you like on the web, then fetch it by slug from inside your agent. Same styles, now callable infrastructure.

Explore the gallery Full MCP reference

......X