Virtual machines
for long-running agents
Sandboxes where agents build and run complex applications,
with persistent storage and programmable networking.
$ curl -fsSL https://superserve.ai/install | shPowering agents at
The cloud was built for requests. Agents need machines that remember
Agents need more than a fresh environment for every task. They need machines they can pause, return to, and branch from as work evolves.
Pause
Keep state while work waits.
Resume
Continue where the agent left off.
Fork
Explore another path from a saved snapshot.
Machines that start really fast
Median time from create() to the first command, with 100 sandboxes launched at once.
- Superserve0.41s
- Cloudflare0.68s1.7×
- Vercel0.72s1.8×
- Modal1.07s2.6×
- Daytona1.09s2.7×
- E2B1.44s3.5×
- Google Cloud Run3.12s7.6×
Hear it from our customers
“We benchmarked several sandbox providers, and Superserve stood out for its speed and reliability. They helped us launch Lindy’s AI teammates in under a month.”
“Superserve lets us give every user a complete game studio with our engine, Playwright, and build tools, without having to run the infrastructure ourselves. They’ve been one of our most dependable partners.”
Everything an agent needs
Build the environment once, keep everything it does, and decide what it can reach.
- Sandboxes
- NVMe storage
- Superserve Metal
Bring any harness, drive it from code
Run Claude Code, Codex, or your own agent loop inside the machine, and control it from TypeScript, Python, MCP, the CLI, or the API.
Enterprise-grade security and controls
Build on open-source SDKs, check our compliance, or run Superserve in your own cloud.
Apache-2.0 SDKs, CLI, and MCP server
The TypeScript and Python SDKs, the CLI, and the MCP server are on GitHub under Apache-2.0.
View on GitHub ComplianceHIPAA compliant, SOC 2 Type II in audit
Policies, our DPA, and audit status are in the trust center.
Visit the trust center Your cloudDeploy in your own VPC
Run Superserve in your AWS, GCP, or Azure account. We're onboarding a small group of early partners.
Book a demoCan I try it for free?
Yes. Start free, no card required. Your agent’s first sandbox runs on a limited free trial. Add a card and choose a plan to continue once the trial credit is used up.
How much does it cost?
Plans start at $20/month, with $20 of usage included. Pro and Enterprise offer bigger sandboxes and higher concurrency. Compare plans and limits →
Is there a time limit on sessions?
Developer and Pro have no fixed session time limit, so agents can work for days or weeks. You can stop or pause a sandbox, or set an auto-pause timeout. Usage is billed for as long as it runs.
What survives a pause?
The whole machine. Files, installed packages, and state are kept, and the agent resumes where it left off.
How is this different from a container?
Every sandbox is its own microVM with its own kernel. No shared kernels or cross-tenant state.
Can we run sandboxes in our own VPC or a customer's VPC?
Yes. We support deploying sandboxes in your own VPC or your customer's VPC. Book a meeting to discuss your setup →
Do agents ever see my API keys?
No. The credentials broker adds them to requests outside the machine, only for hosts you allow.
Give your agent
a computer
$ curl -fsSL https://superserve.ai/install | sh