Quick Answer
ElevenLabs is pitching Creative MCP as one connection for making voice, music, sound effects, images and video from the AI assistant you're already using.
That got my attention.
I spend a lot of time moving between scripts, voiceover, visual assets and editing software. ElevenLabs has now filled in much of the missing setup documentation.
So here's the short version: Creative MCP is now documented for Claude, ChatGPT and Cursor. ElevenLabs publishes hosted setup for Claude Code and Hermes. The FAQ lags, and ChatGPT access varies by plan, so check your tools before spending credits.
Affiliate disclosure: This article contains ElevenLabs affiliate links. If you sign up through one, I may earn a commission at no additional cost to you. That does not change the documentation limits described here.
Try ElevenLabs on one controlled workflow
Start with a short, low-cost test in the documented client and confirm the available tools, permissions, output files and credit use.
Why This Caught My Attention
I'm not looking for another button that promises a finished video. I want to reach a useful first cut faster.
For an explainer, that might mean making temporary narration, trying two visual directions and building a scratch music bed before the real edit. Today, that means bouncing through several apps and poorly named files.
One conversation could make that work less tedious. It still wouldn't choose the best take, fix an edit, mix dialogue or decide whether a shot helps the story.
That's probably where ElevenLabs MCP makes sense for me. Let it get me to the starting line sooner, then finish in tools where I have precise control.
What Is the ElevenLabs MCP?
MCP stands for Model Context Protocol. It's a standard way for an AI assistant to use tools from another service.
The current ElevenLabs MCP product page describes a hosted server that uses OAuth. You sign into an ElevenLabs workspace instead of pasting an API key into a local configuration.
The original open-source server ran locally and used an API key. ElevenLabs' August 22 changelog says it and the MCP player are deprecated. Watch out for older tutorials that still use that setup.
The current product page documents a one-click Claude connection with OAuth and no API key.
Here's Where Things Get Confusing
The product page advertises voice, music, images and video in chat, with more than 50 models behind the same connector. It specifically names Claude, ChatGPT and Cursor, then provides setup instructions for those clients plus Claude Code and Hermes.
The changelog confirms image, video and speech-generation APIs. The creative machinery is there.
The FAQ hasn't caught up: it still names only Claude and Claude Code. Claude's public directory also listed 11 agent-management tools, not the media generators shown in ElevenCreative.
Those gaps still justify checking your tool list, but the new instructions are enough to treat Claude, ChatGPT and Cursor as supported Creative MCP clients. Claude Code and Hermes have setup documentation, not an explicit creative claim.
ElevenLabs now names Claude, ChatGPT and Cursor for voice, music, image and video generation through one connector.
A Creator Workflow I'd Want to Try
Imagine I'm building a three-minute software explainer. I have a working script, but the pacing is still loose.
First, I'd ask for temporary narration. Hearing words against a timeline exposes a sentence that's too long. My ElevenLabs guide for YouTube voiceovers goes deeper on that job.
Next, I might generate two visual directions, a transition sound and a music bed. They don't need to be perfect. They need to answer a few questions: Is the tone right? Does it fit the runtime? Is the music fighting the voice?
For another language, transcription and dubbing could produce a first pass. Names, numbers, pronunciation and the mix still need review. I've covered that in my ElevenLabs localization workflow.
The catch is cost. A casual sentence can launch several billable generations, so I'd want obvious model choices, credit estimates and approval points before trusting a longer request.
Where I Would Take Over
Once the direction works, I'd bring approved files into Premiere Pro, After Effects, Final Cut Pro, DaVinci Resolve or an audio workstation. That's where rough material becomes a finished piece.
My handoff would look something like this:
- Write a tight brief and define the deliverable.
- Generate only the scratch assets I need to make decisions.
- Check the useful ones for accuracy, artifacts, rights and technical quality.
- Name and save the originals somewhere outside the chat.
- Finish the edit, graphics, color, mix, captions and delivery by hand.
AI can get me to the edit faster. It can't watch with a producer's eye, hear music masking a consonant or catch a generated shot that doesn't match the others.
That's useful, but it's not the same thing as finishing a video. If you're comparing voice tools, my best AI voice generators for video creators separates voice quality from the rest of the production job.
How to Use the Hosted ElevenLabs MCP
ElevenLabs documents hosted setup for five clients, but its creative claim specifically names Claude, ChatGPT and Cursor.
Claude and Claude Code
Open ElevenLabs from Claude's connectors directory, choose Connect, sign into the correct workspace and review the permissions. OAuth handles the connection, and you can revoke access later.
ElevenLabs publishes this command for adding the hosted server:
claude mcp add --transport http elevenlabs https://api.elevenlabs.io/v1/mcp
Run /mcp inside Claude Code to finish the OAuth sign-in.
ChatGPT
ElevenLabs provides a Developer Mode and OAuth setup. OpenAI's documentation limits full MCP to Business, Enterprise and Edu; Pro custom apps are read/fetch only. Not every plan can run the same tools.
Cursor
ElevenLabs provides a remote-server configuration and OAuth client ID. Cursor's documentation confirms Streamable HTTP and static OAuth support.
ElevenLabs also publishes a Hermes command, though it doesn't name Hermes for creative generation. Start with one cheap test, keep approval on and confirm the workspace being billed.
The agent-tooling documentation separates the hosted MCP from ElevenLabs' CLI and agent skills. They're different setup paths, not proof that the hosted connector works in an undocumented client.
The hosted connector emphasizes OAuth, workspace-scoped access and permission control.
What I'd Test Before Depending on It
Before using this on client work, I'd check:
- Tool visibility: Do the expected voice, music, image and video tools actually appear in the connected client?
- Model control: Can I choose the model, duration, aspect ratio, voice and output settings?
- Cost visibility: Can I see the likely credit use before a multi-step request runs?
- Iteration: Can I revise one asset without regenerating the entire job?
- Asset handling: Are files named, downloadable and reusable outside the chat?
- Approval: Can read-only and generation actions have different permission rules?
- Quality and rights: Do the outputs survive an edit, mix and delivery review, and are the usage rights clear for this job?
I'd also check consistency. One good image isn't enough if the next four drift in style, and a convincing voice isn't ready if names change between takes.
Is ElevenLabs Creative MCP Worth Trying?
Yes, with a small job and a watchful eye.
Creative MCP is most interesting if you already use ElevenLabs and move between scripts, voices, music, effects and generated visuals. I can see it saving time on rough versions.
I wouldn't build a client process around every advertised feature yet. Tool availability is still inconsistent, generation costs can compound quickly, and this isn't an editing or finishing environment.
For voice quality, dubbing, pricing and commercial-use questions, read my full ElevenLabs review or browse the ElevenLabs creator guide hub.
Check the current ElevenLabs workspace options
Use the current offer only if the documented tools and credit model fit a workflow you can test and review.
Official Sources Checked
- ElevenLabs MCP product page
- ElevenLabs connector in Claude's official directory
- Introducing ElevenLabs MCP
- ElevenCreative changelog for August 17, 2026
- Hosted MCP changelog for August 22, 2026
- ElevenLabs agent-tooling documentation
- OpenAI developer mode and MCP apps documentation
- Cursor MCP documentation
Frequently Asked Questions
What is ElevenLabs Creative MCP?
It's the media-generation side of ElevenLabs' hosted MCP, covering voice, transcription, dubbing, music, sound effects, images and video.
Is ElevenLabs Creative MCP live?
Yes. ElevenLabs now says voice, music, images and video are available through the connector in Claude, ChatGPT and Cursor. Its FAQ is stale, so check the tools exposed in your account.
Which assistants have verified ElevenLabs MCP setup instructions?
The creative section names Claude, ChatGPT and Cursor. Claude Code and Hermes have hosted setup instructions but aren't named in that creative claim.
Does the hosted ElevenLabs MCP require an API key?
No. The hosted MCP uses OAuth sign-in with an ElevenLabs account. The deprecated local open-source server used an API key.
Can ElevenLabs MCP generate music and video?
Yes. ElevenLabs advertises all four in Claude, ChatGPT and Cursor. Tool visibility and ChatGPT plan eligibility can still vary.
Does Creative MCP replace video or audio editing software?
No. It can make source assets, but a finished project still needs editorial judgment, rights review, version control, picture and sound finishing, captions and quality control.
About the Author
Joseph Nilo has been working professionally in all aspects of audio and video production for over twenty years. His day-to-day work finds him working as a video editor, 2D and 3D motion graphics designer, voiceover artist and audio engineer, and colorist for corporate projects and feature films.