Isolated Linux sandboxes in milliseconds — for untrusted code, coding agents, browsers, and autonomous software. One API key, one SDK, one bill.
Ovrin is the agent compute cloud: a self-service platform for provisioning isolated, disposable Linux environments that AI agents and untrusted code can safely run in.
import ovrin
client = ovrin.Client() # reads OVRIN_API_KEY
sandbox = client.sandboxes.create(template="claude-code",
env={"ANTHROPIC_API_KEY": key})
result = sandbox.run('claude "refactor the auth module"')
print(result.stdout)
sandbox.kill()Signup to first sandbox is minutes, and it stays self-service — create an account, mint an API key, and call the API from Python, TypeScript, or plain HTTP.
Hand Claude Code, Codex, Gemini CLI or DeepSeek Harness a whole machine — repo checked out, tools installed, network locked down.
Chrome and Playwright inside the same boundary as the code. The agent clicks, scrapes, uploads and screenshots.
Sandboxes are disposable by design; what the agent learned shouldn't be. Memory persists across sessions.
Run model-generated code safely with stateful interpreter contexts, streaming output, and reproducible environments.
Explore all examples in Examples — they run against any Ovrin sandbox template.
pip install ovrinnpm install ovrin# No install needed — the API is plain REST.
curl https://api.ovrin.app/sandboxes \
-H "Authorization: Bearer $OVRIN_API_KEY"