Join the beta
iPhone
Install TestFlight from the App Store first, then open:
https://testflight.apple.com/join/Uv3YTHRvAndroid closed test
Want to help? Three steps:
Join the Google group
https://groups.google.com/g/parallelsandbox-testersOpen this with the same Google account and tap "Become a tester"
https://play.google.com/apps/testing/com.parallelsandbox.appInstall it from the Play Store on your Android phone
https://play.google.com/store/apps/details?id=com.parallelsandbox.app
If it can't be found right after you join, that's normal. Wait a bit and tap again.
Finished versions, waiting for you to try
Your agent hands the work over. You open it as a user.
ParallelSandbox gives Claude Code, Codex or any MCP agent a real Linux machine to run your project on. When it has a version worth your time it hands it over: a card appears in your app with a picture of that version and one button that opens the product itself. Not a diff, not a log — the running thing.
Download the desktop app
Open your boxes from the same account on macOS or Windows.
A row of front ends waiting for you
By the time a version is handed over it is already running on the machine. The button opens the product on its own URL, so it renders in your browser: native scrolling, no lag, and ten people can open it without touching the box. After handoff, the agent waits for your report within the client’s supported time, then uses your feedback to continue. When a step needs your hands (a login, a second factor, a payment), it asks you to take over the screen and waits for you to hand it back.
What is in a box
A version is a URL you can open
Every service the agent marks as something a person should see gets its own URL. Open it and you are in that version of the product, running in your own browser.
A whole machine, not a container
x86_64 (amd64) Linux with 2 vCPU and 8 GB of memory and 40 GB of working disk per size unit, 1 to 4 units, and its own Docker daemon. docker compose stacks, native builds, anything that needs a real host.
A screen you can see
An Xvfb display with Chromium. Headed browser tests, screenshots, mp4 recordings, and a live view for people.
Plugged into your own dev
Run only the services you changed in the box, under the names the rest of your system uses; everything else stays on your own dev, reached through a connector in your network.
Browser logs for the agent
Pages running in a box report errors through the log SDK; logs_errors returns them with stacks resolved through source maps.
Ready in seconds
Every box is restored from a snapshot and ready in seconds. No time limit while your credits last.
How one job runs
Every step is one MCP tool call from your agent. The last one puts it in your hands.
Start
sandbox_start with a goal and the services your repo exposes. Secrets are injected only when named.
Put the code in
git clone for committed work, sandbox_sync for the uncommitted checkout.
Build and run
docker compose up. Every service name you declared already resolves inside the box; sandbox_wire only matters when unchanged services live at an externalBaseUrl.
Test it yourself first
Run end-to-end tests headed on the screen and fix what breaks. Do not spend a person on a half-finished version.
Hand it over
sandbox_review shows this version in their app immediately and waits for feedback, so the agent can read it and continue in the same conversation.
Ask your AI to set it up
Copy this into Claude Code, Codex or your AI tool. It installs the skill and MCP, then sends you the sign-in link:
Read https:/ / parallelsandbox.com/ skill/ SKILL.md , install the ParallelSandbox skill for me, and complete the connection.Your AI handles installation and setup, and gives you a link when sign-in is needed.
One tool set, over HTTP or stdio
sandbox_start, sandbox_list, sandbox_media, sandbox_exec, sandbox_build, sandbox_sync, sandbox_get, sandbox_pull, sandbox_wire, sandbox_shot, sandbox_scene, sandbox_status, sandbox_say, sandbox_review, sandbox_report, sandbox_takeover, sandbox_device, sandbox_secrets, sandbox_environments, sandbox_publish_version, sandbox_versions, sandbox_draft, sandbox_architecture, sandbox_schema, sandbox_feedback, sandbox_stop, logs_search, logs_errors, logs_tail. The same names over streamable HTTP and the stdio adapter; sandbox_sync and sandbox_pull read and write your local files, so they work only through the adapter.
Logs without copy-paste
Add @parallelsandbox/log to a page. Errors and console output reach your project with release, page path and box id, and the agent queries them directly.
Pay for what runs
Credits, billed by the minute. Free accounts get 1,000 credits that never expire, about 34 hours of a size-1 box. No credit card needed to sign up.
Built for our own agents first
ParallelSandbox runs the AI employees of AStockTeam as one tenant among others: the same MCP, the same boxes, the same takeover flow. What you get is what we use.