sandboxd
Open-source, self-hosted AI app builder — an agent builds real apps in isolated sandboxes on your own server, each live at a preview URL. Self-host in one comma
What is sandboxd?
Open-source, self-hosted AI app builder — an agent builds real apps in isolated sandboxes on your own server, each live at a preview URL. Self-host in one command. MIT.
Overview of sandboxd
What it is
sandboxd is an open-source, self-hosted AI app builder. It runs on your own server and uses an AI coding agent to build real applications inside isolated sandboxes. Each sandbox provides a live preview URL. The entire engine is a single Go program that drives Docker, uses Traefik for URLs, and SQLite for state. It is licensed under MIT.
Key features
- Isolated containers: each app runs in a private, hardened container with its own filesystem and resource limits.
- Live preview URLs: every sandboxed app is accessible at its own URL immediately.
- Sleep and wake: idle sandboxes automatically sleep to conserve server resources and wake on demand.
- Built-in AI agents: supports OpenCode and Claude Code. Credentials are injected via a proxy and never enter the sandbox.
- Checkpoint and revert: every task is checkpointed and revertible.
- Web console and API: an optional no-code web console is available, but everything is also accessible via a
/v1API for headless or scripted use. - Runtime presets: includes presets for React/Vite, Next.js, Node/Express, FastAPI, and Worker, each booting to a live preview.
- Additional features: includes in-browser file editor with diffs, git commit and push, per-app encrypted secrets, snapshots, fork/restore, activity timeline, and per-process logs.
Use cases
sandboxd is suitable for anyone who needs to run many sandboxed environments for other people, such as an AI app builder platform, an agent platform, a coding playground, per-user or per-branch preview environments, or team multi-app hosting. It is not designed for users who only need one or two containers, where a simple shell script or docker run would be simpler.
Getting started
Requires Docker and the Compose plugin, as well as git, on Linux (macOS via Docker Desktop is best-effort). It runs natively on amd64 and arm64. Install in one line: bash curl -fsSL https://raw.githubusercontent.com/tastyeffectco/sandboxd/main/install.sh | bash This command builds the images, starts the stack with the web console, and prints the console URL along with a generated login. The console is available at http://console.localhost and the API at http://127.0.0.1:9090. To run headless (without the console), use SANDBOXD_CONSOLE=0 or --no-console. To upgrade, run ./upgrade.sh.
License and pricing
The software is licensed under MIT and is free and open-source. There is also a managed cloud service called sandboxd Cloud that provides a public URL for your server at $9.99/server/month, with 50% off at launch. Managed installation and management services are also available (BYOC, your compute and data).
Based on the project's public README and repository metadata.