Cartesia in your AI assistant

Q: Should I use the hosted server or run it locally?

Use the hosted server at https://mcp.cartesia.ai/mcp . Sign in with your Cartesia account. Run a local server only if you need to; setup is in the cartesia-mcp README .

Generate speech, transcribe recordings, and clone and manage voices from Claude, Cursor, ChatGPT or Grok.

1
Open Claude Code with the install command ready, or run it in your terminal.
Add to Claude Code
claude mcp add --transport http --scope user cartesia https://mcp.cartesia.ai/mcp
Run /mcp in Claude Code, choose cartesia, and sign in.

Cartesia Skills

Teach your coding agent to write Cartesia code the way the docs describe it. Skills on GitHub

npx skills add cartesia-ai/skills

Work with your recordings and words

Transcribe audio, fix how names are said and find the right voice, all from the assistant you already use.

Transcribe

Turn recordings into text you can search and summarize

Transcribe this standup and list the decisions:standup.wav

Usedspeech_to_text

Launch moves to March 3.
Priya sends the copy to legal by Friday.

Fix a pronunciation

Teach it product names once and every read gets them right

Our product Zyra is said ZEE-ra. Fix it and read the line again.

Usedcreate_pronunciation_dict

Usedtext_to_speech

Find a voice

Search the library by language, accent and style

Find me a warm British narrator.

Usedlist_voices

Three good fits: Theo, Harriet and Callum.

Every tool on the server

Each one runs against your Cartesia account. The cartesia-mcp README has the current list.

Speech

text_to_speechTurns text into speech, with speed, volume, emotion and a pronunciation dictionary. Saves the file and returns a download link valid for 24 hours.
speech_to_textTranscribes audio in batch or streaming mode.

Voices

list_voicesLists voices, filtered by language, search term, gender and more.
get_voiceReturns the details of one voice.
clone_voiceClones a voice from a recording.
update_voiceChanges a cloned voice's name or description.
delete_voiceDeletes a cloned voice.
localize_voiceAdapts a voice to another language or dialect.

Accents and models

list_accentsLists the catalog accents and their details.
list_modelsLists the text to speech and speech to text models you can call.
add_voice_accentsAdds catalog accents to an instant voice clone.
delete_voice_accentRemoves a catalog accent from an instant voice clone.

Pronunciation

list_pronunciation_dictsLists your pronunciation dictionaries.
create_pronunciation_dictCreates a pronunciation dictionary.
get_pronunciation_dictReturns one pronunciation dictionary.
update_pronunciation_dictChanges a pronunciation dictionary.
delete_pronunciation_dictDeletes a pronunciation dictionary.

Account

download_fileFetches a Cartesia file by ID and returns a download link.
get_credit_usageReports credit usage over time. Admin only.

Cartesia MCP server tools
Group	Tool	What it does
Speech	`text_to_speech`	Turns text into speech, with speed, volume, emotion and a pronunciation dictionary. Saves the file and returns a download link valid for 24 hours.
Speech	`speech_to_text`	Transcribes audio in batch or streaming mode.
Voices	`list_voices`	Lists voices, filtered by language, search term, gender and more.
	`get_voice`	Returns the details of one voice.
	`clone_voice`	Clones a voice from a recording.
	`update_voice`	Changes a cloned voice's name or description.
	`delete_voice`	Deletes a cloned voice.
	`localize_voice`	Adapts a voice to another language or dialect.
Accents and models	`list_accents`	Lists the catalog accents and their details.
	`list_models`	Lists the text to speech and speech to text models you can call.
	`add_voice_accents`	Adds catalog accents to an instant voice clone.
	`delete_voice_accent`	Removes a catalog accent from an instant voice clone.
Pronunciation	`list_pronunciation_dicts`	Lists your pronunciation dictionaries.
	`create_pronunciation_dict`	Creates a pronunciation dictionary.
	`get_pronunciation_dict`	Returns one pronunciation dictionary.
	`update_pronunciation_dict`	Changes a pronunciation dictionary.
	`delete_pronunciation_dict`	Deletes a pronunciation dictionary.
Account	`download_file`	Fetches a Cartesia file by ID and returns a download link.
Account	`get_credit_usage`	Reports credit usage over time. Admin only.

MCP server FAQs

What is MCP?

The Model Context Protocol is an open standard that lets AI assistants call outside tools. Cartesia's MCP server gives your assistant tools for speech, transcription and voices, so you can ask for them in a chat instead of writing API calls.

Which AI assistants work with it?

Cartesia documents setup for Claude Code, Cursor, ChatGPT and Grok. Most clients should use the hosted server at https://mcp.cartesia.ai/mcp. For other setups, see the cartesia-mcp README.

What does it cost?

Tool calls use your Cartesia credits, like any other API call. Generating speech or transcribing audio draws from your plan. See pricing for plans and credits.

Do I need an API key?

Not for the hosted server. You sign in with your Cartesia account, and a Cartesia MCP API key is created for your organization if it doesn't have one. A local install needs your own key; see the cartesia-mcp README.

Can it build voice agents?

Not through MCP yet. The tools cover speech, transcription, voices and pronunciation. To build and deploy a voice agent, use Cartesia agents.

Should I use the hosted server or run it locally?

Use the hosted server at https://mcp.cartesia.ai/mcp. Sign in with your Cartesia account. Run a local server only if you need to; setup is in the cartesia-mcp README.

Get started today

Talk to an expert.

Connect with a member of our team and learn how Cartesia can help you build world-class voice experiences.

Contact Sales

Start building.

Access our models via API and bring a voice agent into production in minutes.

Try Cartesia