Take Friday Off in your AI assistant

What your assistant can do with it

• Search every guide, case study, question and note on takefridayoff.com, and get the passages that answer the question, each linked to its exact section (tfo_search).
• Read any page in full, as clean Markdown with its sources and the date it was last reviewed (tfo_get_page).
• Check a product's pathway. Medicine, food or cosmetic? The check applies the rules on our pathways page, in their order, and quotes the rule that decides it, with what to confirm next (tfo_pathway_check). In assistants that show MCP Apps, the answer is a card you can try variations in. It's indicative, not legal or regulatory advice.
• Find case studies by format, ingredient, pathway or client (tfo_case_studies).
• Take the readiness check, the same four questions as our partner page, and see whether we're a fit (tfo_readiness_check).
• Ask for a call with Gabriel Perera, our founder, and get a link to pick a time. The assistant has to ask you first (tfo_request_call).

It also answers questions about who we are, the Launch Engine and our founder, and offers ready-made prompts, such as a product brief interview.

Skills

The server also carries step-by-step workflows, called skills, that an assistant can follow: a launch brief for a product idea, the medicine, food or cosmetic check, and a TGA listing plan. Assistants that support skills over MCP load them from the server. You can also add them to your own assistant with npx skills add https://takefridayoff.com.

Connect it

The address is the same everywhere: https://takefridayoff.com/mcp. It needs no account and no authentication.

Claude

Add a custom connector with the address above (Connectors, then Add custom connector), and turn it on for a chat. In Claude Code:

claude mcp add --transport http takefridayoff \
  https://takefridayoff.com/mcp

ChatGPT

Turn on developer mode in ChatGPT's settings, create an app or connector with the address above and no authentication, then choose it for a chat.

Cursor and VS Code

In Cursor, .cursor/mcp.json:

{
  "mcpServers": {
    "takefridayoff": { "url": "https://takefridayoff.com/mcp" }
  }
}

In VS Code, .vscode/mcp.json:

{
  "servers": {
    "takefridayoff": { "type": "http", "url": "https://takefridayoff.com/mcp" }
  }
}

On your own machine

For assistants that start their own servers, the npm package runs the same server over stdio. In Claude Desktop's claude_desktop_config.json:

{
  "mcpServers": {
    "takefridayoff": { "command": "npx", "args": ["-y", "takefridayoff", "serve"] }
  }
}

From a terminal

The same tools, without an assistant. It needs Node.js 18.17 or later and nothing else.

npx takefridayoff search TGA listing fees
npx takefridayoff page /guides/tga-approval
npx takefridayoff work capsule
npx takefridayoff fit --role product --stage scale --drag capacity --timeline 6
npx takefridayoff book --open

npx takefridayoff help lists every command; --json gives output a script can read.

Try asking

• "How much does it cost to list a supplement with the TGA, and how long does it take?"
• "Is a collagen powder for skin a food or a medicine in Australia?"
• "When can an SPF moisturiser be sold as a cosmetic?"
• "Show me Take Friday Off's case studies for capsule products."

What we record

When an assistant uses the server, we record which tool it used, what it asked for (a search, a page, a product description), the name of the assistant and the outcome, to learn what people look for. We don't record IP addresses, and a call request reaches us only to answer it. The command-line tool reads the site's public content and sends nothing else unless you ask for a call.

For developers

• Protocol. Streamable HTTP, one JSON response per request, no sessions: MCP 2026-07-28 (server/discover) and the initialize handshake of 2024-11-05 to 2025-11-25.
• Discovery. The server card is at /mcp/server-card (and /.well-known/mcp/server-card.json); in the official MCP Registry the server is com.takefridayoff/mcp.
• The pages, for machines. llms.txt and llms-full.txt, and every page as Markdown at its address with .md on the end.
• REST API. The same tools over plain HTTP, with no key: GET /api/search?query=TGA+listing+fees, /api/page?path=/guides/tga-approval, /api/case-studies, /api/faq and the checks. It's described in OpenAPI 3.1 at /api/openapi.json, and listed with the MCP server at /.well-known/api-catalog.
• In the browser. An AI agent working in Chrome gets the same tools through WebMCP, and our contact form fills itself in for you to check and send.