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

# Overview

> Ways to integrate Supertone — official Python and TypeScript SDKs, a terminal CLI, and an MCP server for AI agents.

Supertone offers several ways to call the Text-to-Speech API beyond raw REST. Pick the one that matches how you work: SDKs for application code, the CLI for terminals and scripts, and the MCP server for AI agents. They all wrap the same REST API and stay in sync with the OpenAPI specification.

<CardGroup cols={2}>
  <Card title="Python SDK" icon="python" href="/en/docs/developer-tools/python">
    `pip install supertone` — sync and async clients, parallel auto-chunking.
  </Card>

  <Card title="TypeScript SDK" icon="js" href="/en/docs/developer-tools/typescript">
    `npm add @supertone/supertone` — promise-based, ESM/CJS, Node and Bun.
  </Card>

  <Card title="CLI" icon="terminal" href="/en/docs/developer-tools/cli">
    `pip install supertone-cli` — synthesize, stream, batch, and manage voices from the terminal.
  </Card>

  <Card title="MCP" icon="robot" href="/en/docs/developer-tools/mcp">
    `uvx supertone-mcp` — composable tools for Claude, Cursor, and other AI agents.
  </Card>
</CardGroup>

## Which one should I use?

| If you want to…                                           | Use                                                                                               | Audience                      |
| --------------------------------------------------------- | ------------------------------------------------------------------------------------------------- | ----------------------------- |
| Call Supertone from Python or Node code                   | [Python](/en/docs/developer-tools/python) / [TypeScript](/en/docs/developer-tools/typescript) SDK | Application developers        |
| Generate or batch audio from the shell, scripts, or CI    | [CLI](/en/docs/developer-tools/cli)                                                               | DevOps, content/scripting     |
| Let an AI agent discover voices and synthesize speech     | [MCP server](/en/docs/developer-tools/mcp)                                                        | Claude / Cursor / agent users |
| Use a language without an SDK, or control the wire format | [REST API](/en/api-reference/introduction)                                                        | Any runtime                   |

## What each tool gives you

**SDKs (Python & TypeScript)** — Idiomatic clients for application code. They handle authentication, expose type-safe models and enums, stream audio chunks, auto-chunk long text (split at 300 characters and merged transparently), raise typed errors keyed to HTTP status codes, and support configurable retries with backoff. Each SDK page has a full feature table.

**CLI** — The API from your terminal and scripts. Synthesize from a string, file, or stdin; stream to your speakers; batch a whole directory; manage and clone voices; check usage and predict duration. `--format json` output is pipe-friendly for automation, and conventional exit codes let scripts branch on failures.

**MCP server** — Composable tools that let an AI agent (Claude, Cursor, and other MCP clients) drive Supertone. The agent can discover and preview voices, estimate cost, clone voices, and synthesize speech — chaining these into multi-step voice workflows on its own.

## When to use the REST API directly

The REST API is available for languages and runtimes where an SDK isn't published — Go, Java, PHP, Rust, embedded systems — or when you want maximum control over the wire format. The endpoints are documented in the [API Reference](/en/api-reference/introduction); examples in that section include `curl`, JavaScript, Go, Python, PHP, and Java.

If you call REST directly, you'll need to implement chunking, retries, and streaming consumption yourself.