Skip to content
Jan Arvin Mito / Amito

Clearer instructions.
Better starting points.

Give your AI assistant a practical starting point for an interface. Find requirements, check what’s missing, and turn decisions into a specification you can build from.

Use with AI

https://amito.dev/mcp

Remote MCP · Streamable HTTP · No sign-in or API key

  1. In your assistant’s settings, add a custom MCP server or remote connector.
  2. Name it “Amito”, enter the URL above, and choose no authentication.
  3. Enable it in your conversation and ask for an interface specification.

Setup guides: ChatGPT · Claude · Gemini · Perplexity.

Custom connections depend on your account and workspace settings. These links explain each platform’s setup; Amito is not automatically installed or listed in their directories.

What it offers

  • Interface specifications

    Starting requirements for forms, tables, onboarding, and dashboards. Each includes loading, empty, error, keyboard, and responsive behavior.

  • A check for missing decisions

    Give the assistant your draft. It can check for missing fields and obvious placeholders, then return a checklist of decisions to make.

  • Public information about my work

    Read my frontend focus, current contribution at Dupe, and personal projects, with links to the original pages.

Browse the specifications →

Try asking

Use Amito to find the form specification. Help me write a signup form spec, then check which decisions are still missing.

The check confirms that decisions are documented. It does not judge the design, inspect an implementation, or certify accessibility. Review and test the finished interface.

For developers

The server exposes three read-only tools: search, fetch, and validate_design_spec. Search returns document IDs and source URLs. Fetch a specification to get the exact decision keys, then pass your draft to the validator.

{
  "pattern": "form",
  "decisions": {
    "purpose": "Help a new customer create an account.",
    "primaryAction": "Create account; confirm success and continue to setup."
  }
}

Example arguments for validate_design_spec. This incomplete draft returns the remaining decisions. Requests are limited to 64 KiB, with up to 30 decisions of 2,000 characters each.

Data and privacy

The server reads a fixed collection of public documents. It cannot open your files, browse other websites, change a project, or make purchases. Draft decisions are processed in memory and are not saved or logged by the MCP application.

Send only the interface behavior needed for the check. Your AI platform and hosting infrastructure may keep their own conversation or access logs. Keep credentials, personal data, and confidential client material out of your request.