mcpField
O
onboarding

onboarding

github.com/elevenlabsDevelopment5231.3k installsUpdated 2d ago
Repository

About onboarding

A setup skill for the ElevenLabs CLI that guides app developers through authentication, API-key configuration and SDK integration. It provides step-by-step instructions and safeguards for handling credentials and project files.

Key features

  1. 01Initialize setup from an app folder
  2. 02Follow JSON-guided setup steps
  3. 03Configure SDK key loading
  4. 04Handle approvals and user input
  5. 05Verify setup with onboard check

Use cases

  1. 01Set up ElevenLabs in a project
  2. 02Add SDK access to an existing app
  3. 03Resolve a missing API key configuration

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/elevenlabs/skills --skill onboarding

After installing, type/onboardingin Claude Code to confirm it loaded.

Details

From the project README

ElevenLabs Onboarding

The ElevenLabs CLI runs the whole setup and tells you each step. From the folder of the app that will use the key (in a monorepo, that app's folder), run:

npx -y @elevenlabs/cli@latest onboard init

Every onboard command prints JSON. Act on it, and keep going until next is done:

  • next: what to do now, spelled out in instruction.
  • run: the exact next command. Run it as given.
  • say: what to tell the user. Tell them as written.
  • details: facts for the next step, such as how the app loads the key and the SDK install command.
  • rules: rules for the whole setup. Follow them.

What each next means:

  • run: run run.
  • wait: the approval page is open. Show say (it has the link), then run run again right away, without ending your turn. Whenever you give the user the link, copy it whole, exactly as printed: a shortened link doesn't work.
  • ask_what_to_build: tell the user say, which asks what they want to build or whether to just add the SDK, and wait for their answer before writing any code. Ask it even if they said earlier, and don't pick a product for them. Once they answer, run run.
  • write_code: write what they asked for, as instruction says: the SDK alone, or the integration, with the skill that fits from details.fetch_skills when there is one. If it needs something made in ElevenLabs first, like an agent, make it yourself with the key (load the env file in the command, never print it), not a script for the user to run. Load the key the way details.load_key says, then run run.
  • load_key: the code already uses the SDK, so there is nothing to build. Make sure the app loads the key the way details.load_key says, change nothing else, then run run.
  • fix: do what say describes, then run run. If it needs the user's shell, settings or secrets, ask them instead, as for ask_user.
  • ask_user: ask the user what say says and wait for their answer, then run run, unless they said no.
  • offer_test: say what you set up and anything the user still has to do themselves, like installing the packages you added. Then end your reply with the question in say. Run run only if they say yes.
  • done: tell the user say, then anything they still have to do themselves, in one or two concrete steps.

Rules

  • Never open, print or cat the env file; onboard check verifies it.
  • 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.
  • Never run git clean or delete untracked files: the key is in one.
  • Don't write or change any code until a reply's next is write_code or load_key.
  • Put anything you download only to read, like a package or an archive, in a temp folder, never in the project.
  • Don't run the app (or its start script) or make speech to try it until the user says yes to the test request. API calls that what they asked for needs, like creating an agent, are fine. Once the code is written, run onboard check before you finish: it verifies the key and the code.

No npx

If npx isn't available (Node.js isn't installed), install the CLI (or update it if it's already installed: brew upgrade elevenlabs, or run the installer again), then run elevenlabs onboard init instead. Everything after that is the same.

  • macOS or Linux: brew install elevenlabs/tap/elevenlabs, or curl --proto '=https' --tlsv1.2 -LsSf https://github.com/elevenlabs/cli/releases/latest/download/elevenlabs-cli-installer.sh | sh
  • Windows: powershell -ExecutionPolicy ByPass -c "irm https://github.com/elevenlabs/cli/releases/latest/download/elevenlabs-cli-installer.ps1 | iex"

If elevenlabs isn't found after the installer ran, use ~/.cargo/bin/elevenlabs in its place, here and in every run the CLI gives.

Comments

Markdown supported