ToastBuddy

Remote MCP

ToastBuddy for OpenAI

Same public MCP as Muse and Grok. Paste this exact URL. No API key.

Paste this exact URL

https://toastbuddy.com/api/mcp

Streamable HTTP. No bearer token. No login.

Discovery

https://toastbuddy.com/.well-known/mcp.json

https://toastbuddy.com/.well-known/mcp.json

Sample agent prompt

Find a warm birthday speech example, explain what makes it work, then make cue cards and give me the editable deck link. Help me replace invented details with my own.

Search the library, read one example, then use its cleared library_slug with make_cue_cards.

Connect in OpenAI or Cursor

  • Add remote MCP → https://toastbuddy.com/api/mcp, no key.
  • Discovery: https://toastbuddy.com/.well-known/mcp.json
  • In Cursor: add remote MCP → https://toastbuddy.com/api/mcp, no key.
  • Same six public tools: draft_toast, make_cue_cards, shorten_toast, export_cards, search_speech_examples, get_speech_example.

Tools

  • draft_toast — Occasion, relationship, tone, length, stories, jokes. About a cent per call.
  • make_cue_cards — Turn your draft or a cleared library example into cards with an exact editable web-deck link. No LLM spend.
  • shorten_toast — Trim to a target time. Deterministic. No LLM spend.
  • export_cards — Markdown or a short-lived PDF link. Deterministic. No LLM spend.
  • search_speech_examples — Find public examples by topic, occasion, tone or kind. Up to ten results with sources, rights and speaking time. No private data or LLM spend.
  • get_speech_example — Read one public example, its rights and what makes it work. Complete cleared text; otherwise published excerpts and analysis only. No LLM spend.

From inspiration to your own cue cards

Read an example, borrow its structure, and use your own true memories. Every result includes its rights: complete cleared text is available; other works remain excerpts, analysis and the publisher’s player. Recording rights are separate.

Browse speech examples →Find quotes by topic →Make printable cue cards →Learn from memorable speeches →

Connect this MCP in a host that supports remote HTTP tools. These connection pages are not a claim of marketplace listing, client certification or automatic discovery by every assistant.

Live demo · make_cue_cards · no LLM

Canned cue cards for Alex

Same deterministic splitter the MCP runs. 7 cards from a 60s birthday toast. spent_llm: false.

  1. Card 1 · opening

    Friends, I want to raise a glass to Alex.

  2. Card 2 · body

    As your friend, the thing I keep coming back to is how they show up.

  3. Card 3 · body

    The night they drove across town just to sit with you says it all.

  4. Card 4 · body

    That is the person we came to celebrate.

  5. Card 5 · body

    Warm, steady, and somehow still the first to laugh.

  6. Card 6 · body

    Here is to another year of that same generous energy.

  7. Card 7 · closing

    To Alex.

Caps

  • A single draft_toast is about a cent. The ~$8 figure is the daily global budget, not the price of one draft.
  • 20 LLM drafts per IP per day, 400 globally, ~$8/day global budget.
  • Library search/read, cue cards, shortening, and export do not call a model. Library reads: 60 requests per IP per minute per running instance; search returns 5 results by default, maximum 10.
  • Editable deck links are limited to 16,000 characters. If a deck is too large, the complete cards are returned with a clear copy/paste fallback; nothing is silently truncated.

Usage shows up as utm_source=mcp / utm_campaign=openai so you can verify the connect in PostHog.

Support: support@toastbuddy.com

Also: /for/muse · /for/grok