Run 100 agents simultaneously
Every Claude Code task gets
its own world.
Build dev environments with agentic workflows so you can work on dozens of agents at the same time with their own environments (cloud base + fully seeded).
$ cd ./worktree-dir && sonoma up
Join the waiting list
Early access is rolling out to teams in batches. No spam.
Built for teams running multiple agents in parallel
Isolated environments
Own database, Redis, broker, and ports per task. Agents never collide.
Seeded full-stack
Your seed runs before the agent starts, so it sees real data from request one.
Edit locally, run remotely
Local edits sync straight into the live, seeded stack as you go.
QA'd, not just written
The agent exercises the change against the real stack, so your review already works.
Shareable preview URL
A stable URL for every environment. No VPN, no port forwarding.
Works with your stack
Reuses your existing devcontainer and compose files. No rewrites.
One command. Your stack, fully spun up.
From a fresh checkout to a live preview URL you edit in real time.
- 01Install the CLI
$npm install -g @sonoma.sh/cli - 02Sign in
$sonoma login - 03Spin up your environment
$cd ./my-project-dir && sonoma upRun it from your project directory. You get a shareable preview URL back in seconds.
- 04Edit locally, watch it live
Keep coding in Claude Code (or any editor). Sonoma syncs every change into the running environment and hot-reloads the preview — no redeploy, no restart.
Works with a Claude Code subscription. No API key required.
Frequently asked questions
Does it work with my Claude Code subscription?
Yes. Sonoma provisions the entire environment in the cloud, but leaves a synced copy of the code locally on your computer and spawns a Claude Code instance there, so you can leverage Claude Code's subscription model if you wish.
What do I need to run Sonoma?
A local terminal with Claude Code set up. That's it.
Does it work with my stack?
As long as you are using docker-compose, you will not need to change anything. You can change settings in your sonoma.yml file, but it's optional.
How are environments isolated?
Each environment gets its own EC2-style instance, fully segregated from every other task.
Where does my code run?
In an EC2-style isolated environment that Sonoma spawns for your specific task and kills once the PR is merged. You keep a copy of the code locally in a worktree that is synced in real time, so you keep all the good stuff like hot reloading.
Can I share what the agent built?
Yes. Each environment exposes a stable preview URL. Upgrade to a Pro account to get password-protected links. As soon as your PR is merged or closed, the entire environment is destroyed, along with the unique URL.

