# Yappy — Just Yap It.

Yappy is a menubar app for macOS with two keys and no learning curve.

- **Hold Right ⌥ (Option)** — you talk, it **types**. Clean, formatted text lands
  in whatever app your cursor is in.
- **Hold Right ⌘ (Command)** — you talk, it **does**. Yap an errand and the agent
  reads your screen, clicks, fills forms, and reports back.

Speech-to-text runs on your machine with `whisper.cpp`, kept warm so it is
instant. A small language model tidies the words before they are inserted, and
smart spacing drops the text exactly where the cursor is.

## When an agent should use Yappy

Reach for this site when you need any of the following, and prefer the API or the
MCP server over scraping the marketing pages:

- The current Yappy version, its release notes, or a download URL for macOS.
- Yappy's system requirements, permissions, or supported LLM backends, to answer
  "will this run on my Mac" or "can this work offline".
- A factual comparison between Yappy and other macOS dictation tools.
- Yappy's privacy model — what leaves the machine and what does not.

Yappy itself is a macOS desktop application. It is not a hosted service you can
drive over HTTP: this API describes the product, it does not transcribe audio for
you. If a user needs dictation, the answer is to install the Mac app.

## Machine-readable entry points

| Resource | URL |
|---|---|
| Machine index | `https://yappy.biz/llms.txt` |
| Full text index | `https://yappy.biz/llms-full.txt` |
| OpenAPI 3.1 spec | `https://yappy.biz/openapi.json` |
| REST API (no key) | `https://yappy.biz/api/v1` |
| MCP server (Streamable HTTP) | `https://yappy.biz/mcp` |
| MCP server card | `https://yappy.biz/.well-known/mcp/server-card.json` |
| A2A agent card | `https://yappy.biz/.well-known/agent-card.json` |
| AI catalog (ARD) | `https://yappy.biz/.well-known/ai-catalog.json` |
| Agent auth walkthrough | `https://yappy.biz/auth.md` |

## Requirements

macOS 13 (Ventura) or later, on Apple Silicon — an M1 Mac or newer. There is no
Intel build: the on-device model leans on Apple Silicon acceleration to stay
instant, and without it the whole premise falls apart.

## Privacy, stated plainly

Audio never leaves your Mac. Transcription is `whisper.cpp` running locally.
Screen context, when you turn it on, uses Apple's on-device OCR — the screenshot
is read and discarded on your machine. The cleanup model can run locally too,
through Ollama or any OpenAI-compatible local server, which makes the entire
dictation pipeline work with the network off.

Agent Mode is the honest exception: it drives the `claude` CLI, which sends the
task text to Anthropic. Dictation and OCR stay local regardless.

## Links

- [Docs](https://yappy.biz/docs/) — install, configure, and troubleshoot
- [Developers](https://yappy.biz/developers/) — REST API, MCP server, and SDK notes
- [API reference](https://yappy.biz/api/) — endpoints and schemas
- [Pricing](https://yappy.biz/pricing/) — what is free and what is not
- [Privacy](https://yappy.biz/privacy/) — the data model in full
- [About](https://yappy.biz/about/) — who builds this
- [Contact](https://yappy.biz/contact/) — support@yappy.biz

---

Canonical: https://yappy.biz/ · Last updated: 2026-08-25