Skip to main content
CreativoAI

CreativoAI for AI agents

This page is for AI agents and assistants (and the people who build them). It says what CreativoAI offers, how to read it, and what it does not offer.

The tools

  • Image to prompt: Upload an image and get a structured text-to-image prompt describing what is visible: subject, setting, composition, lighting, palette and style. Input: an image.
  • Prompt enhancer: Paste a rough text-to-image prompt and get a clearer version with useful detail, a concise variant, and three alternatives, keeping your constraints. Input: text.
  • Image style analyzer: Break down the style of an image — composition, lighting, texture, mood and colours — and get tips and a reusable style prompt. Input: an image.
  • Image caption generator: Get three distinct captions, a short version, calls to action and hashtag ideas for Instagram, LinkedIn, Pinterest or general use. Input: an image.
  • Image alt-text generator: Generate alternative text for an image based on its purpose and page context, including when to leave alt empty. Input: an image.
  • Image keyword generator: Get a title, short description, ordered and deduplicated keywords and category ideas for stock photos, blog images or product libraries. Input: an image.

Reading the site

  • Every public page can be requested as markdown: send the header Accept: text/markdown. HTML stays the default.
  • Discovery files: sitemap, AI catalog, API catalog, OpenAPI description and an agent skills index.
  • A public, read-only MCP server at /mcp (server card): list the tools, read each tool's fields and limits, and read the guides. It needs no authentication and cannot run a tool.
  • In a browser that supports WebMCP, each page offers tools to list the tools, open one, and fill its form fields. The person still completes the human-verification check and presses the run button.
  • robots.txt carries a Content Signals line: search and use as AI input are allowed; training is not.

What is not offered

  • No public API. The API behind the tool pages is for the website itself: same-origin requests only, an anonymous cookie, a small number of free daily credits per visitor, and a human-verification token for every run.
  • No accounts, API keys, OAuth or agent-to-agent endpoint. The MCP server is read-only information; it cannot run the AI tools.
  • Results are AI-generated and can contain mistakes; they should be reviewed by a person.

Questions

Use the contact page.