Introducing the ElevenLabs CLI v1

Today we are introducing v1 of the ElevenLabs CLI that brings the ElevenLabs API directly into your terminal.

Coding agents already live in the terminal where they run commands, read results, and evenorder lunch. The ElevenLabs CLI was designed to be agents first, providing well documented commands and structured JSON results that are easy to parse and chain. It ships with a --dry-run mode to preview any operation before it touches your workspace.

The CLI also brings agents-as-code to ElevenAgents. Pull every agent in your workspace into local config files, edit them like any other part of your codebase, preview the exact diff, and push to production by hand or through your coding agent.

Install on macOS

Windows

cURL

The entire platform from one interface

Every ElevenLabs API endpoint published in our OpenAPI spec is available as subcommands:

Agents-as-code

Agents-as-code treats agent configuration the way you treat application code. You work on files on disk, and the CLI keeps them in sync with your workspace. The whole workflow fits in three commands.

The same workflow extends to production. Branches keep development and production separate - a dev branch can use a test phone number and a cheaper LLM while production runs the real number and a frontier model. And for new projects, elevenlabs agents init scaffolds everything, with starter templates for common configurations like customer support.

With a handful of agents, the dashboard is manageable. However, when managing a fleet of agents spread across hundreds of customer orgs or bulk migrating agents from another platform, files and version control are the only workflow that scales.

Built for coding agents and developers

An agent using a CLI fails differently than a person does. It cannot answer an interactive prompt, it retries constantly, and it learns from whatever the error message gives it. The CLI is designed around this reality.

--help is for people, --schema is for agents. Every API command prints a machine-readable contract like typed inputs, where each one goes, what's required, and the exact shape of the response. This lets the agent construct a valid call without guessing.

Errors also give the agent something to work with. For agents, errors come back as JSON it can parse. When a human user is detected via TTY, a plain error message is shown instead. Both include instructions on how to fix the error.

Given the invalid command:

Agents would get the following error message:

Whereas human users see this error that is more legible and includes a hint on how to fix the problem:

Batteries are included with skills

--help and --schema are the CLI's built-in manual. But skills go further and teach your coding agent complete workflows like how to build an agent, generate speech, or transcribe a call.

Install them into your project:

Agents already live in the terminal. Now all of ElevenLabs does too. Install the CLI, hand your agent the skills, and see what it ships. Get started with docs.