> ## Documentation Index
> Fetch the complete documentation index at: https://docs.withhopper.com/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP connector

Hopper's MCP server gives an AI app Hopper's voice and tools: it can speak, transcribe, look up models and prices, return tested integration code, and review a voice agent for latency. Speech, transcription and the lookup tools need no account.

```text theme={null}
https://withhopper.com/mcp
```

Streamable HTTP. Sign-in, when a tool needs it, is OAuth with your Hopper account.

## Add it

```bash Claude Code theme={null}
claude mcp add --transport http hopper https://withhopper.com/mcp
```

```bash Codex theme={null}
codex mcp add hopper --url https://withhopper.com/mcp
```

```json Cursor (.cursor/mcp.json) theme={null}
{
  "mcpServers": {
    "hopper": { "url": "https://withhopper.com/mcp" }
  }
}
```

* **Claude** (claude.ai and the desktop app): **Settings → Connectors → Add custom connector**, then paste the URL.
* **ChatGPT**: turn on **Developer mode** in **Settings → Security and login**, then add the URL at **chatgpt.com/plugins → +**.
* **Coding agents**: the [Hopper plugin](https://github.com/hopper-inc/plugins) adds this connector plus skills that set Hopper up in a voice-agent project and benchmark it. In Claude Code: `claude plugin marketplace add https://withhopper.com/marketplace.json`, then `claude plugin install hopper@hopper`.

## Tools

| Tool | Sign-in | What it does |
| :- | :- | :- |
| `speak` | No | Text to speech, `text` and optional `voice`. Returns 24 kHz mono WAV and a playable link (10 minutes, 3 plays); ChatGPT and other MCP Apps hosts show an audio player |
| `transcribe` | No | English speech to text with word timestamps, up to 5 minutes per call. WAV, MP3, M4A, Ogg/Opus, FLAC or WebM: a file attached in ChatGPT (`audio_file`), `audio_url` (public https, up to 25 MB) or `audio_base64` (up to 3 MB) |
| `list_voices` | No | Voices ready for speech: id, name, language, gender, description |
| `list_models` | No | Models, context length, prices and availability, as the API bills them |
| `get_integration_guide` | No | Tested code that switches a voice agent's LLM to Hopper. `framework`: `livekit`, `pipecat`, `vapi` or `openai-sdk` |
| `review_voice_agent_config` | No | Checks a `system_prompt`, optional `tools` and `client_code` for what delays the first token, with fixes |
| `get_account` | Yes | Credits, spend, request and token counts, and API keys (names and first characters only) |
| `create_api_key` | Yes | Creates a key on your account, optional `name`, and returns a one-time link to copy it in the console. The key never appears in the chat |

## Free use and sign-in

* **ChatGPT, not signed in**: each ChatGPT user gets their own allowance for speech and transcription. Near its limit the tools say so; at the limit, ChatGPT offers to connect a Hopper account. The daily limits below also apply.
* **Other apps, not signed in**: speech and transcription work under shared limits: 10 calls a minute, about a minute of speech per call, and 10 minutes of speech and 30 minutes of transcription a day.
* **Signed in**: speech and transcription bill your own credits; new accounts start with free credit (see [Models](https://docs.withhopper.com/models.md)).

Signing in asks for these scopes: `account:read` (`get_account`), `keys:write` (`create_api_key`) and `audio` (speech and transcription on your credits). Your app opens Hopper's sign-in page the first time a tool needs one.

## What it doesn't do

The connector can't change account settings, delete or reveal existing keys, or buy credits; those are in the [console](https://withhopper.com/console). It doesn't call the LLM: use the [API](https://docs.withhopper.com/llm.md) for completions.

## Data

The connector receives the arguments of the tool you call. Text you speak and audio you transcribe aren't stored. See the [privacy policy](https://withhopper.com/privacy).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.