ElevenLabs skills are ElevenLabs’ own free set of eleven instruction folders that teach a coding agent, such as Claude Code, to use ElevenLabs’ voice service: text-to-speech, transcription, sound effects, music, voice changing, voice cleanup, dubbing and voice agents. The skills cost nothing, but every one needs an ElevenLabs API key, and the text and audio you work with are sent to ElevenLabs. The older local ElevenLabs MCP server is archived; a hosted one, which you sign in to, has replaced it.
- What: eleven official skills from ElevenLabs, MIT licence, for Claude Code and other agents that read skills.
- Needs: an ElevenLabs account and API key. The voice work is charged to your ElevenLabs account; the skills are free.
- Data: your text and audio go to ElevenLabs, whose privacy policy says it trains its models on personal information unless you opt out.
- The old MCP server: archived. Its replacement is hosted by ElevenLabs and uses a sign-in instead of a pasted key.
The skills are a recipe book; ElevenLabs is the kitchen. The recipe tells your AI agent exactly how to order a voiceover or a sound effect from ElevenLabs, but the cooking happens on ElevenLabs’ computers, on your account. That is why every recipe needs your key, and why what you send leaves your machine.
| Question | ElevenLabs skills | Hosted MCP server | Old local MCP server |
|---|---|---|---|
| What it is | Instructions the agent reads, then writes code that calls ElevenLabs | A connector run by ElevenLabs that your AI app plugs into | A program you ran on your own computer |
| Status | Maintained | Current | Archived, “no longer actively maintained” |
| How you sign in | API key in your project’s settings file | A sign-in page (OAuth); no key copied | API key pasted into a config file |
| Main use | Voiceover, transcripts, sound, music, dubbing, voice agents | Managing voice agents, plus speech from text | Speech, cloning, transcription and more |
Each row is sourced in the sections below (read 19 Oct 2026).
What are ElevenLabs skills?
The plain point: they are ElevenLabs’ official how-to notes for AI coding agents, one per ElevenLabs product.
A skill is a folder of written instructions an AI agent reads only when a task needs it. ElevenLabs describes its set as “Agent skills for ElevenLabs developer products” that “can be used with any compatible AI coding assistant.”1 The README lists eleven:
- text-to-speech (turning text into a spoken voice) and speech-to-text (transcripts with timestamps);
- sound-effects and music, made from text descriptions;
- voice-changer (re-voicing a recording in a different voice) and voice-isolator (removing background noise);
- dubbing into other languages “while preserving the original speakers’ voices”;
- agents and speech-engine, for building voice assistants that talk back in real time;
- onboarding and setup-api-key, which handle signing in and the key.
The repo is MIT-licensed (free to use, including in paid work), was created in January 2026 and had about 500 GitHub stars on 19 Oct 2026.2 Stars show attention, not quality.3
Note the word “developer”. The skills do not press buttons in an app for you. They teach the agent to write and run small programs that call ElevenLabs’ API (the door one program uses to ask another for work). The text-to-speech skill, for example, shows Python and JavaScript code that sends text and saves an MP3 file.4 You do not have to read the code, but you are running it.
The skills are free know-how. The voice work itself happens at ElevenLabs, on your account.
What needs an API key?
The plain point: all eleven skills do. The hosted connector uses a sign-in instead.
The README is direct: “All skills require an ElevenLabs API key in ELEVENLABS_API_KEY.” An API key is a password that lets a program use a paid online service on your account. The onboarding skill gets one for you: it opens an approval page in your browser, then “puts one in your project’s env file” (a hidden settings file in the project folder).5 Its rules tell the agent to keep the key out of the chat and out of the code:
- “Never ask for an API key in chat, and never accept one pasted there. If one appears, tell the user to rotate it on the API keys page.”
- “Never put the key in code; the SDK reads
ELEVENLABS_API_KEY.” - “Don’t run the app (or its start script) or make speech to try it until the user says yes to the test request.”6
The voice work itself is charged to your ElevenLabs account. The dubbing skill, for instance, warns that editing a source transcript after a language is done means a “(charged) regeneration”.7 For plans, see ElevenLabs’ pricing page; prices change, and that page is the only current one.
What happened to the ElevenLabs MCP server?
The plain point: the local one you installed yourself is retired. ElevenLabs now runs one for you, aimed mainly at its voice agents.
An MCP server is a connector that lets an AI app use another service directly. The old local ElevenLabs server is archived, and its README now opens: “This local MCP server is deprecated in favor of the ElevenLabs hosted MCP server.” It adds that the hosted server “authenticates with OAuth, so no API keys are copied into your client” and that “This repository is no longer actively maintained.”8 OAuth is the “Sign in with” kind of login, where you approve access on the service’s own page.
ElevenLabs’ docs say the hosted server “is published in the Claude Desktop directory”. Its tools mostly manage voice agents: you can “Create new agents by describing what you want”. It can also “Generate speech audio from text, returned as a short-lived download link.”9 The page mentions no music, sound effects, cloning or dubbing tools, so for those, use the skills.
The same docs carry two warnings worth knowing before you connect: “Updating an agent through the MCP server publishes the change directly”, and “Deleting an agent is destructive.” They advise: “Review tool calls in your MCP client before approving them.”10
For voiceover, sound and dubbing in a project, use the skills. For managing ElevenLabs voice agents from a chat, use the hosted connector. Skip the archived local server.
Where do your text and audio go?
The plain point: to ElevenLabs. Read its privacy settings before you send anything private.
Every skill “Requires internet access” because the work happens on ElevenLabs’ servers (skill files, above). ElevenLabs’ privacy policy, updated 9 October 2026, says: “We use your personal information to research, build, train and improve our AI models.” Its list of data used for that includes “Inputs & Outputs”. You can opt out in the “Data use” menu, under “Terms and Privacy” in your account, and the policy notes the opt-out only covers training after you switch it.11 Where you use ElevenLabs through an employer or other business customer, the policy says “the business customer’s contract with us, not this Policy, governs how we use the information.”11
In this site’s reading, three habits follow. Switch the opt-out on before your first job if you do not want your material used for training. Do not send recordings of other people, client scripts or meetings without permission. And keep the env file that holds your key out of anything you share or upload.
Can you use someone else’s voice?
The plain point: not without their clear permission. ElevenLabs’ rules forbid it.
The voice-changer and dubbing skills can make speech sound like another voice, including a cloned one. ElevenLabs’ Prohibited Use Policy bans “Cloning, synthesizing, or impersonating the voice, identity, or likeness of any living individual” “without their explicit and informed consent or other valid authorization”, and says “You may not submit voice recordings of anyone under 18 to create or train a voice model.”12 The voice-changer skill itself separates the two jobs: it converts a recording into an existing voice and says cloning a new voice from a sample “is voice cloning (IVC/PVC), a separate feature”.13 Read voice cloning before you try either.
How do you install ElevenLabs skills?
The plain point: one command adds them; the onboarding skill then walks you through the key. This site installed and ran nothing.
Three safety steps first: install only from the official elevenlabs account on GitHub, not a look-alike; skim the SKILL.md files you plan to use; and decide on the training opt-out before you send real material. The install command is from the README; the general steps for each app are on how to install AI skills.
The command needs Node.js, which brings npx, a tool that downloads and runs JavaScript programs.14 For the hosted connector instead, ElevenLabs’ docs point to the connector directory in Claude Desktop (source above).
What is a good first test?
The plain point: one short voiceover of your own words, in a stock voice, in an empty folder.
Listen to the whole file before you use it. For how to write text that sounds natural (pauses, emphasis, tags), see the ElevenLabs cheat sheet.
What ElevenLabs skills cannot do
The plain point: they make sound, not pictures, and they only work as well as your account allows.
- No video. To put a voiceover under motion graphics, pair them with a video tool such as HyperFrames or Remotion.
- No offline mode. Every skill needs internet access and the key (skill files, above).
- Written for builders. The skills lean towards adding voice to an app. For a single voiceover, ElevenLabs’ own website may be simpler (this site’s reading).
Questions people ask
Are ElevenLabs skills free?
The skills are free under the MIT licence, but each needs an ElevenLabs API key and uses your plan’s credits (sources above). See ElevenLabs’ pricing page for plans.
Do they only work with Claude?
No. ElevenLabs says they “can be used with any compatible AI coding assistant” that follows the Agent Skills standard (source above).
Is the ElevenLabs MCP server dead?
The local one is archived and unmaintained. ElevenLabs now runs a hosted MCP server you connect to by signing in (sources above).
Does ElevenLabs train on what I send?
Its privacy policy says it uses personal information, including inputs and outputs, to train its models, and that you can opt out in your account’s “Data use” menu (source above). Business contracts can differ.
Where to go next
- “What other video and audio skills are there?” See AI video skills, sorted by job.
- “Skill, MCP or plugin: which is which?” Read skills vs MCP vs plugins.
- “How do I check a skill before installing?” Use the pre-install checklist.
- “How do I get good voice results?” See the ElevenLabs cheat sheet, prompting music and voice and sound in AI video.
- “What do these words mean?” Text-to-speech, speech-to-text, API and voice cloning.
- “Back to the section.” All pages on AI skills.
QUICK CHECK · 3 QUESTIONS
Pick an answer to see if you got it. Nothing is recorded.
1. What do all ElevenLabs skills need before they can work?
- A paid Claude plan with voice mode
- A microphone plugged into your computer
- An ElevenLabs API key
Show the answer
Answer: An ElevenLabs API key. The README says all skills require an ElevenLabs API key; the onboarding skill can set one up for you.
2. What happened to the old local ElevenLabs MCP server?
- It was archived and replaced by a hosted server you sign in to
- It became the only official way to use ElevenLabs
- It was merged into the text-to-speech skill
Show the answer
Answer: It was archived and replaced by a hosted server you sign in to. The page says the local server is archived and unmaintained, and ElevenLabs now runs a hosted MCP server that uses a sign-in instead of a pasted key.
3. What does the page say about the text and audio you send?
- It never leaves your computer
- It goes to ElevenLabs, which may train on it unless you opt out
- It is deleted the moment the file is made
Show the answer
Answer: It goes to ElevenLabs, which may train on it unless you opt out. The work happens on ElevenLabs’ servers, and its privacy policy says it trains its models on personal information, with an opt-out in the Data use menu.
- 1ElevenLabs, elevenlabs/skills README, read at source 19 Oct 2026“Agent skills for ElevenLabs developer products. These skills follow the Agent Skills specification and can be used with any compatible AI coding assistant.”; licence “MIT”; eleven skills listed ↑
- 2GitHub API, elevenlabs/skills repository data, read at source 19 Oct 2026“created_at”: “2026-01-24”, “stargazers_count”: 495, “pushed_at”: “2026-10-18”, licence “MIT”, “archived”: false ↑
- 3arXiv, Six Million (Suspected) Fake Stars in GitHub, v2 6 Sep 2025, read at source 13 Oct 2026 (as cited on this site’s HyperFrames page): “the star count is the most widely used popularity signal, but it is also at risk of being artificially inflated (i.e., faked)”↑
- 4ElevenLabs, elevenlabs/skills
text-to-speech/SKILL.md, read at source 19 Oct 2026“Convert text to speech using ElevenLabs voice AI. Use when generating audio from text, creating voiceovers, building voice apps, or synthesizing speech in 90+ languages.”; “compatibility: Requires internet access and an ElevenLabs API key (ELEVENLABS_API_KEY).”; the Quick Start code writes “output.mp3” ↑ - 5ElevenLabs, elevenlabs/skills README, “Configuration”, read at source 19 Oct 2026“All skills require an ElevenLabs API key in ELEVENLABS_API_KEY. The onboarding skill puts one in your project’s env file for you.” ↑
- 6ElevenLabs, elevenlabs/skills
onboarding/SKILL.md, “Rules”, read at source 19 Oct 2026“Never ask for an API key in chat, and never accept one pasted there. If one appears, tell the user to rotate it on the API keys page.”; “Never put the key in code; the SDK reads ELEVENLABS_API_KEY.”; “Don’t run the app (or its start script) or make speech to try it until the user says yes to the test request.”; the flow has “the user leaves the conversation once, to approve a browser page” ↑ - 7ElevenLabs, elevenlabs/skills
dubbing/SKILL.md, read at source 19 Oct 2026“editing the source after a language completes marks it stale and requires a (charged) regeneration.” ↑ - 8ElevenLabs, elevenlabs-mcp README and repository data, read at source 19 Oct 2026“This local MCP server is deprecated in favor of the ElevenLabs hosted MCP server.”; “It authenticates with OAuth, so no API keys are copied into your client.”; “This repository is no longer actively maintained.”; “archived”: true, last push “2026-08-20”; the README describes the old server as one that lets MCP clients “generate speech, clone voices, transcribe audio, and more” ↑
- 9ElevenLabs Documentation, Hosted MCP server, no date shown, read at source 19 Oct 2026“The hosted MCP server is published in the Claude Desktop directory.”; “Create new agents by describing what you want”; “Generate speech audio from text, returned as a short-lived download link.”; “Authentication uses OAuth.” ↑
- 10ElevenLabs Documentation, Hosted MCP server, “Permissions” and “Security”, read at source 19 Oct 2026“Updating an agent through the MCP server publishes the change directly.”; “Deleting an agent is destructive.”; “Review tool calls in your MCP client before approving them.”; “The assistant can only perform actions covered by the permissions you approve.” ↑
- 11ElevenLabs, Privacy Policy, updated 9 Oct 2026, read at source 19 Oct 2026“We use your personal information to research, build, train and improve our AI models.”; the purpose “Research and develop our AI models” lists “Inputs & Outputs”; “You can opt out of our use of your personal information at any time by navigating to the ‘Data use’ menu in the ‘Terms and Privacy’ section of your ElevenLabs account.”; “The opt-out only applies to and affects use of your personal information for training after you opt-out.”; “We won’t keep data we generate about your voice for longer than 3 years after your last interaction with us” ↑a“This Policy does not cover personal information we collect when you use ElevenLabs services through one of our business customers”; “the business customer’s contract with us, not this Policy, governs how we use the information.” ↑b
- 12ElevenLabs, Prohibited Use Policy, last updated 9 Oct 2026, read at source 12 Oct 2026 (as cited on this site’s talking photo page): “Cloning, synthesizing, or impersonating the voice, identity, or likeness of any living individual” “without their explicit and informed consent or other valid authorization”, and “You may not submit voice recordings of anyone under 18 to create or train a voice model.”↑
- 13ElevenLabs, elevenlabs/skills
voice-changer/SKILL.md, read at source 19 Oct 2026“Do not use this skill to create or clone a new voice from a voice sample — that is voice cloning (IVC/PVC), a separate feature; this skill only converts an existing recording into an existing voice_id.” ↑ - 14ElevenLabs, elevenlabs/skills README, “Installation”, read at source 19 Oct 2026“npx skills add elevenlabs/skills”;
onboarding/SKILL.md: “Requires internet access to elevenlabs.io and api.elevenlabs.io, and Node.js for npx” ↑
Read at source 19 Oct 2026: ElevenLabs, elevenlabs/skills (README and the SKILL.md files for text-to-speech, onboarding, dubbing and voice-changer) and repository data; ElevenLabs, elevenlabs-mcp README and repository data; ElevenLabs Documentation, Hosted MCP server; ElevenLabs, Privacy Policy. Re-used from this site with their read dates: ElevenLabs, Prohibited Use Policy (12 Oct 2026); arXiv, Six Million (Suspected) Fake Stars in GitHub (13 Oct 2026). Pricing link: ElevenLabs pricing (no prices shown on this page).
Nothing on this page was installed or run by this site: every behaviour is described from ElevenLabs’ own files and pages. The recipe-book comparison, the example prompt and the advice marked “this site’s reading” are this site’s own, not claims from these sources.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK