> This is a page from the ElevenLabs documentation. For an index of the documentation, fetch https://elevenlabs.io/docs/llms.txt.

# ElevenCreative MCP

## Overview

The ElevenCreative MCP connects AI assistants such as Claude, ChatGPT, and Cursor to your ElevenCreative workspace through the [Model Context Protocol (MCP)](https://modelcontextprotocol.io/). Your assistant can generate speech, music, sound effects, images, and video, transcribe and dub audio, and design new voices. Every result is saved to an editable [Flow](/docs/eleven-creative/products/flows) that you can open in the app.

There is nothing to install. You sign in with your ElevenLabs account through OAuth, so no API keys are copied into the client, and generations use the credits of the workspace you connect.

## Server URLs

| Server                                 | URL                                                            | Includes                                                                                                                |
| -------------------------------------- | -------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------- |
| ElevenCreative                         | `https://api.elevenlabs.io/v1/mcp-creative`                    | Every creative tool.                                                                                                    |
| ElevenCreative in the Claude directory | `https://api.elevenlabs.io/v1/mcp-creative-claude-marketplace` | Speech, voices, sound effects, and transcription. Image, video, and music generation are not available in this version. |

### Data residency regions

The default server URL connects to the global environment. If your workspace is in an isolated [data residency](/docs/overview/administration/data-residency) environment, connect with your region's URL instead:

| Region    | MCP server URL                                           |
| --------- | -------------------------------------------------------- |
| EU        | `https://api.eu.residency.elevenlabs.io/v1/mcp-creative` |
| India     | `https://api.in.residency.elevenlabs.io/v1/mcp-creative` |
| Singapore | `https://api.sg.residency.elevenlabs.io/v1/mcp-creative` |

> **Warning**
>
> Isolated environments are separate workspaces with separate accounts. When the OAuth flow starts,
> sign in with your account for that environment, not your elevenlabs.io account.

## ElevenCreative or ElevenLabs

The [ElevenLabs MCP](/docs/eleven-agents/operate/hosted-mcp) (`https://api.elevenlabs.io/v1/mcp`) includes every ElevenCreative tool plus the ElevenAgents tools for building and managing voice agents. Connect ElevenCreative if you only need creative work. Connect ElevenLabs if you also build agents.

You don't need both. Connecting both gives your assistant two copies of the same creative tools.

> **Note**
>
> For the agent tools, setup, and permissions, see the [ElevenLabs MCP](/docs/eleven-agents/operate/hosted-mcp) page.

## Connect

### Claude

* **From the directory:** open **Settings** > **Connectors**, search for ElevenCreative, select **Connect**, and sign in with your ElevenLabs account.
* **Full version:** open **Settings** > **Connectors** and select **Add custom connector**. Name it ElevenCreative, enter `https://api.elevenlabs.io/v1/mcp-creative`, and select **Add**. Then sign in, choose a workspace, review the permissions, and select **Authorize**. The custom connector includes image, video, and music generation.

### ChatGPT

Open the apps directory, search for ElevenCreative, select **Connect**, and sign in with your ElevenLabs account.

### Cursor

Install the ElevenCreative plugin from the Cursor Marketplace, or add the server to your `mcp.json`:

**`mcp.json`**

```json title="mcp.json"
{
  "mcpServers": {
    "elevencreative": {
      "url": "https://api.elevenlabs.io/v1/mcp-creative"
    }
  }
}
```

### Other clients

Point any MCP client that supports remote servers at `https://api.elevenlabs.io/v1/mcp-creative` and complete the OAuth flow. Clients that support hosted client metadata (CIMD) need no separate client registration.

## What you can make

* **Text to speech:** voiceovers and narration in any workspace or voice library voice. Speech returns two takes of each line.
* **Voice design:** describe a voice, hear previews, and save the one you choose to your workspace.
* **Speech to text:** transcripts with speaker labels and timestamps in 99 languages.
* **Dubbing:** translate a track into another language while keeping the speaker's voice and delivery.
* **Music:** tracks in any genre, with or without vocals.
* **Sound effects:** effects from a short prompt, with several variations to choose from.
* **Images:** generate images and edit them, from text or from a reference.
* **Video:** generate video from text or animate an image, then add a voiceover or lipsync.
* **Brand kits:** apply your logos, colors, fonts, and voices to every generation.
* **References:** upload files from your computer or attach a file or link the assistant can reach.

## How Flows work

Every generation lands on a Flow, an editable canvas that holds the prompts, references, and results. The assistant returns a link to it, and Claude also shows it inline. To chain steps, such as a video with a voiceover or a lipsync from generated speech, keep them on the same Flow.

Images and video return several variations by default. Tell the assistant which one you want, for example the second image, and it carries that one into the next step. You can open any Flow in ElevenCreative to change a prompt, rerun a node, or finish the piece in Studio.

## Tools

| Group                 | Tools                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                         |
| --------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Generate              | `creative_generate_speech`, `creative_generate_image`, `creative_edit_image`, `creative_generate_video`, `creative_transcribe_audio`, `creative_generate_in_flow` (music, sound effects, dubbing, lipsync, and every other Flow node)                                                                                                                                                                                                                                                                                         |
| Voices                | `creative_list_voices`, `creative_design_voice`, `creative_get_voice_design_previews`, `creative_save_designed_voice`                                                                                                                                                                                                                                                                                                                                                                                                         |
| Flows                 | `creative_create_flow`, `creative_list_flows`, `creative_get_flow`, `creative_get_flow_node`, `creative_add_flow_node`, `creative_add_flow_asset_node`, `creative_update_node`, `creative_update_flow_nodes`, `creative_connect_flow_nodes`, `creative_disconnect_flow_nodes`, `creative_duplicate_flow_nodes`, `creative_delete_flow_nodes`, `creative_set_flow_port`, `creative_expose_flow_parameter`, `creative_run_flow_nodes`, `creative_get_flow_run_status`, `creative_show_flow_results`, `creative_get_generations` |
| References and assets | `creative_upload_flow_reference`, `creative_attach_reference_file`, `creative_get_available_assets`                                                                                                                                                                                                                                                                                                                                                                                                                           |
| Brand kits            | `creative_list_brand_kits`, `creative_get_brand_kit`, `creative_create_brand_kit`, `creative_create_brand_kit_from_website`, `creative_update_brand_kit`, `creative_delete_brand_kit`, `creative_add_brand_kit_item`, `creative_update_brand_kit_item`, `creative_delete_brand_kit_item`                                                                                                                                                                                                                                      |
| Models                | `creative_get_flow_node_types`, `creative_get_model_guide`, `creative_get_model_schema`                                                                                                                                                                                                                                                                                                                                                                                                                                       |

## Example prompts

* "A product shot of a ceramic mug on a marble counter, morning light, 16:9."
* "Turn the second image into a five-second video with a slow push-in."
* "Add a 20-second warm British voiceover to the same flow."
* "Design a gruff sea captain voice and give me three previews."
* "Make a 60-second lo-fi music bed with no vocals."
* "Transcribe this interview and list the speakers."
* "Dub this clip into Spanish and keep the original speaker's voice."

## Generated files

Files come back as signed links on `elevenlabsusercontent.com`, which is kept separate from elevenlabs.io for security. Links expire about two hours after delivery. To get a new one, ask the assistant to fetch the generation again or open the Flow in the app. If your network or client restricts media domains, allow `elevenlabsusercontent.com`.

## Credits and permissions

* Generations spend credits the same way they do in the app.
* Don't ask the assistant to retry a generation that is still running. Each retry starts and charges a new generation.
* The consent screen lists the permissions the connector asks for, such as Text to Speech, Flows, and Image & Video Generation. The assistant can only act within those permissions and your workspace role.
* Revoke access at any time from your MCP client or your ElevenLabs account settings. In Claude, you can choose which tools run freely and which need approval, and workspace admins can set this for the whole organization.

## Troubleshooting

#### Images, video, or music don't run in Claude

The directory version doesn't include them. Add the custom connector with
`https://api.elevenlabs.io/v1/mcp-creative`.

#### A download link stopped working

Links expire after about two hours. Ask the assistant to fetch the generation again.

#### The assistant picks the wrong connector

If you have both ElevenLabs and ElevenCreative connected, remove the one you don't need.