LiveLLM.
Log inGet started

A home lab,
without the hum.

Machines, apps, databases and browsers you can reach from anywhere. Set them up yourself, or hand the job to your AI agent.

Start a workspace
Rather run it on your own hardware? It's open source.
Your agent
➜Read https://docs.p4c.live-llm.com/SKILL.md and set up LiveLLM for me
✓Skill installed. Open the link I sent you and press Allow.
✓Signed in to acme.
➜Put github.com/you/my-site online with a Postgres database
✓Created main-db and my-site, built from main (a1b2c3d).
✓Live at https://my-site-http-acme.cloud.p4c.live-llm.com
acmeAll running
my-siteApp from GitHub
main-dbPostgreSQL

Reach everything

Every app gets an HTTPS address, every machine SSH and a screen in your browser.

Let your agent drive

It builds, deploys and fixes things while you watch, within the limits you set.

Or do it as code

The console, Terraform, the API and the command line all do the same things.

What people run on it

All recipes →

Ship a side project

Deploy from your Git repo or an image, add a database, and get an HTTPS address in minutes.

See the recipe →

Keep a dev box in the cloud

An Ubuntu, Debian or Fedora machine you reach over SSH with your own keys, from any laptop.

See the recipe →

Run Windows anywhere

Windows 11 in your browser or over Remote Desktop, or Windows Server with a command line. No Windows PC needed.

See the recipe →

Automate the web

Real Chromium browsers you drive with Playwright or Puppeteer, and can watch live.

See the recipe →

Give your AI agent a computer

Let Claude, Codex or Cursor use browsers and machines while you watch. You decide what it may do.

See the recipe →

Run an app of several services

A web service, a cache and a worker on one private network, started in the right order.

See the recipe →

Ask your agent to set it up

Paste one line into your agent. It learns LiveLLM, asks you to let it in, and gets to work.

➜Read https://docs.p4c.live-llm.com/SKILL.md and set up LiveLLM for me

One line to start

Paste the prompt into Claude Code, Codex, Cursor or any agent. It installs the skill and signs in.

Skills to read, tools to run

The skill teaches your agent how LiveLLM works. The connector gives chat apps like Claude.ai the tools to act.

You stay in charge

Press Allow once, choose how far it may go, and sign it out from the Agents page whenever you like.

Connector, for chat apps

Paste this address into Claude.ai's connectors, or any app that takes one.

https://api.p4c.live-llm.com/mcp
Setup guide →

Skill, installed by hand

Rather install it yourself than paste the prompt? Pick your agent.

Install with
mkdir -p ~/.claude/skills
curl -fsSL https://codeload.github.com/qalby-tech/livellm_cloud_skills/tar.gz/main \
  | tar -xz -C ~/.claude/skills --strip-components=2 \
    livellm_cloud_skills-main/skills/livellm-cloud
Into Claude Code’s skills folder. Change the folder for your agent.
The skill on GitHub →

Its job doesn't end at deploy

LiveLLM keeps each thing's uptime, alerts on CPU, memory and disk, and logs every change. Your agent reads the same, so when something breaks you can just ask it what happened.

→“Why is my-site down?” It reads the logs and the restarts.
→“Restart it, and give it more memory.”
→“What did you change this week?” It's all in Activity.
acme · what happened today
my-site is back upIt ran out of memory and restarted. Your agent raised its memory to 1Gi.
main-db disk is 85% fullYour agent asks before growing it.
dev-box stoppedIts stop time ran out. The disk is kept.

Questions, answered

Your own space on LiveLLM: the machines, apps, databases and browsers you create, on their own private network, within your plan’s CPU, memory and disk.

Your rack is empty.
Fill it.

Start a workspace