Superserve and Petrova.
Superserve rents the computer your agent runs on. Petrova is the account your agent calls. They do different things, so you could use both.
How Superserve works
Cloud computers (sandboxes) for AI agents.
- You or your agent install its command-line tool, or use its SDKs or API. The agent shows you a link and a code, and you approve it. No pasted keys.
- Each agent gets its own small virtual machine, made from a saved template.
- The agent works inside: shell commands, files, Docker, a browser.
- You can pause a machine, wake it later with everything still in memory, or copy it to try several paths.
- You set which sites it can reach. A secrets feature keeps real keys outside the machine. You pay per second for what it uses.
How Petrova works
The account your agent calls. Today this all runs on test data.
- Your agent keeps running where it runs today. Petrova does not run it.
- You change two settings: the base URL and the key. Anthropic and OpenAI formats both work.
- Before each call, Petrova holds the most that call could cost.
- A supplier answers. If one fails before the first word, the call moves to another.
- You are charged what the call used, plus a 5.5% fee. The rest comes back. Never above the hold.
How they fit together
- Your agent runs inside a Superserve sandbox.
- It calls Petrova for models and tools, using the base URL and a Petrova key.
- The Petrova key can be stored as a Superserve custom secret, limited to the Petrova address. The real key then stays out of the machine.
- Superserve bills you for the machine. Petrova holds and caps what the agent spends on calls.
No partnership exists. We have not tested the two together, and Petrova is not live.
| Superserve | Petrova | |
|---|---|---|
| Live today | Yes | No. Preview on test data |
| Runs the machine your agent lives on | Yes | No. Petrova sells no compute |
| Pause, resume and copy an agent | Yes | No |
| Keeps real keys out of the agent's reach | Yes, with stand-in keys | No. A key has scopes and caps, but the agent holds it |
| Sign-in flow for agents, with no pasted keys | Yes, live | Planned. Mock only |
| Sells or routes model calls | No. You pay model companies yourself | Yes (test data) |
| Caps on model spend | Not found | Account, project, key, agent, session (test data) |
| Charge never above a hold set before the call | Not found | Yes (test data) |
| Paid tools and memory on one balance | Not found | Yes (test data) |
Green marks the stronger answer in a row, red the weaker. "Test data" means Petrova's code does it in tests, not as a service. "Not found" means we did not find it on their public pages on the check date. It may still exist.
Where Superserve is ahead
- Live today
- Runs the machine your agent lives on
- Pause, resume and copy an agent
- Keeps real keys out of the agent's reach
- Sign-in flow for agents, with no pasted keys
What we could not check
- Its speed and scale claims.
- Its compliance status.
- The size of its free trial.
Read on 2 November 2026, on their public pages: superserve.ai · superserve.ai/pricing · superserve.ai/SKILL.md · docs.superserve.ai · its secrets post. Things may have changed since. Each claim of theirs is their own statement.