ElevenLabs launches CLI v1 for terminal-based voice agents

ElevenLabs has launched CLI v1, a command-line tool that exposes its entire voice agent platform and OpenAPI specification as terminal subcommands. Available via Homebrew, Scoop, or curl, the tool introduces an agents-as-code workflow. Developers can now pull their entire voice agent configuration into local files, edit them, preview changes with a dry-run diff, and push them back to production. This transition from web dashboards to local configuration files allows teams to manage voice agents using standard version control and CI/CD pipelines.

A key innovation of the CLI is its dual-mode error handling, which detects whether a human or an AI agent is executing a command by checking the TTY. While human users receive readable hints, AI coding assistants get structured JSON outputs containing precise error codes and fix instructions. Additionally, every command supports a --schema flag that outputs a machine-readable contract of inputs and response shapes. For example, running elevenlabs text-to-speech convert --schema provides a JSON object detailing every parameter, its type, and required fields, enabling AI agents to construct valid API calls without human intervention.

The CLI also includes an elevenlabs generate-skills command, which writes workflow instructions directly into a project. These skills teach external AI assistants, such as Claude Code, how to build agents, generate speech, or transcribe calls. This command-line tool is designed to complement ElevenLabs' hosted Model Context Protocol (MCP) server. While the MCP server is optimized for conversational management, the CLI is built for bulk migrations, multi-tenant SaaS configurations, and automated deployments. By treating voice agents as code, the release helps developers scale their voice infrastructure from a few prototypes to hundreds of customer-specific deployments.