LiveKit CLI
LiveKit CLI installs skills and Docs MCP for coding agents, and its debugger lets one talk to a voice bot.
- 384 ★ on the repository
- Last commit
Extended access
Why this level
- The tool uses LiveKit project keys and manages rooms and agent deployments.
- Installing skills and the Docs MCP changes agent configuration on the machine.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/livekit-cliDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The tool uses LiveKit project keys and manages rooms and agent deployments.
- Installing skills and the Docs MCP changes agent configuration on the machine.
This is third-party code. Review the repository files before installing.
What it does
`lk` is LiveKit's command-line tool: it bootstraps apps from templates, creates access tokens, manages rooms and egresses, joins a room as a participant and runs load tests. For coding agents it has three parts. `lk skills install` copies the official LiveKit skills into Claude Code, Codex, Cursor and seven other agents and adds the Docs MCP to their config. `lk docs` searches documentation and code from the terminal. `lk agent debugger` starts your voice agent and converses with it through `say` and `chat-history`.
Who it is for. Developers of LiveKit voice and video agents who want Claude Code or Cursor to test its own changes to the agent.
Good fit when
- A voice agent is edited in Claude Code and every change should be tested by conversation without a microphone.
- Agents in a new project need LiveKit skills and the Docs MCP from one command.
- A room token is issued from the console without opening the LiveKit Cloud dashboard.
- A deployment is loaded with test participants before launch.
Not a fit when
- The project does not use LiveKit: the tool is tied to its API.
- The agent test must go through a real microphone and phone line.
- Skills and the Docs MCP are unwanted, and extra prompts from the tool should be avoided.
Example request
Start my voice agent with `lk agent debugger`, ask what it can do and show the tool calls.Limitations
The tool works with LiveKit Cloud or your own LiveKit server, so room and token commands need project keys. Building from source needs Git LFS plus CGO, PortAudio and ALSA headers on Linux. If a coding agent without LiveKit skills is found, `lk` offers once to install them globally; turn it off with `LK_NO_SKILLS_PROMPT=1`. The debugger's `--audio` mode speaks turns through LiveKit Inference TTS.
How to disable. Run `lk skills remove` to drop the skills and delete the Docs MCP from the agent configuration. Uninstall the tool the way it was installed.
FAQ
How do I install LiveKit skills into chosen agents?
Run `lk skills install --agent claude-code --agent codex`. Without flags it uses agents detected on the machine, and `--global` installs for the user.
How do I update the skills?
With `lk skills update`. Skills you edited locally are not overwritten without `--force`.
How can an agent test a voice agent?
From the project directory run `lk agent debugger start`, then `say` with a line and `logs --last 40` if something looks wrong. `--json` gives script-friendly output.
How do I search the docs?
`lk docs search "voice agents"` searches the documentation and `lk docs code-search` searches code across LiveKit repositories.
How do I turn off the skills install offer?
Set `LK_NO_SKILLS_PROMPT=1`. In CI, with `--quiet` or `--json`, the offer never appears.
Discussion
No questions or reviews yet.
Related
CLI toolsSeven LiveKit skills for voice agents in Python and Node.js: docs, debugging, tests, simulations and operations
Kode CLI is a terminal coding agent that reads AGENTS.md and CLAUDE.md, connects MCP servers and works with many models.
Over twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
A set of 116 skills, 44 agents and safety hooks for Claude Code, Cursor, Codex, Gemini CLI and nine more editors
Black-box auditor runs a released CLI, catches side writes to disk and gives agents a command index with a readiness score.
Apify CLI is the Apify platform console that creates, runs and deploys Actors and prints a ready SKILL.md for coding agents.
README in short
The debugger is built for coding agents: each turn prints tool calls, errors and the reply, and the exit code shows whether the turn passed. A session ends itself after 30 idle minutes.