The SpeechifyAI CLI is here: drive the API from your terminal

There’s now a command line for Speechify. Install it, log in with your API key, and you can synthesize speech and browse voices without writing a line of integration code. It’s the companion to the developer console: same API, from your terminal instead of a browser tab.

npm i -g @speechify/cli
speechify login
speechify say "hello from the command line"

That’s the whole getting-started. Three lines to your first audio.

Log in once with your API key

speechify login takes a Speechify API key, validates it against the server, and stores it in your operating system’s keychain. You do it once. Every command after that uses the stored key.

speechify login    # paste your API key
speechify logout   # forget the stored key

Grab an API key from the console. The key already carries its own workspace and project scope, so there’s nothing else to select. If you’d rather not store anything, pass --api-key on any command or set $SPEECHIFY_API_KEY.

What ships in this release

• say turns text into speech, with a voice, a format, and a file to write to.
• voices list browses the catalogue with filters; voices get shows one voice in full: its models, locales, tags, and preview URLs.
• api is a raw, gh-style passthrough to any endpoint, for the things the higher-level commands don’t cover.
• mcp relays Speechify’s hosted MCP server over stdio, and mcp install wires that relay straight into your AI client.

Say something

speechify say "Text to speak" \
  --voice henry \
  --format wav \
  --out narration.wav \
  --play

echo "from a pipe" | speechify say -

--voice defaults to george. --format is one of wav, mp3, ogg, aac, or pcm (default mp3). --out defaults to ./speech.<format>, and - streams the audio to stdout. --play plays it after synthesis, and speechify say - reads the text from stdin.

Find a voice

speechify voices list --locale en --gender female --search warm
speechify voices get scott

voices list filters by --locale (en for all English, en-US for exact), --gender (male, female, not_specified), and --search, which matches against a voice’s id, name, and tags.

Built for agents, not just people

The CLI is agent-friendly by design. Every command takes --json for machine-readable output, --agent-friendly for JSON plus context an agent can act on, and --no-input so it never blocks on a prompt and returns a needs-input spec instead.

And mcp install hands Speechify to your coding agent directly:

speechify mcp install --client claude-code cursor
speechify mcp install --all   # every detected client

It writes the relay into the MCP config for Claude Code, Cursor, Claude Desktop, Windsurf, or VS Code, so the agent drives Speechify through the same tool you do, no separate integration.

Why a CLI at all

Because not everything needs an app. Checking how a voice sounds, generating a one-off clip, wiring Speechify into a shell script or a CI step: these are all faster from a terminal than from code or a dashboard. The CLI is for the quick things, and a foundation for the scripted ones.

The full command reference lives in the CLI repo. Next up, a deeper walkthrough of say: reading text aloud from your terminal end to end.

FAQ

How do I install the SpeechifyAI CLI?

Run npm i -g @speechify/cli, then speechify login and paste an API key from the console. After that, speechify say "…" synthesizes speech from the terminal.

Does the CLI use an API key?

Yes. speechify login validates a Speechify API key and stores it in your OS keychain. You can also pass --api-key per command or set $SPEECHIFY_API_KEY. The key carries its own workspace and project scope.

What can it do?

In this release: say for text to speech, voices list and voices get, a raw api passthrough, and mcp / mcp install for AI agents.

Can an AI agent use it?

Yes. speechify mcp install wires Speechify’s hosted MCP server into Claude Code, Cursor, Claude Desktop, Windsurf, or VS Code. Every command also supports --json, --agent-friendly, and --no-input.