Routant and Petrova.
Routant decides what an agent may do in a company's systems. Petrova decides what it may spend. Neither runs the agent, so you could use both.
How Routant works
A checkpoint between AI agents and private systems.
- An agent company keeps its agent where it runs today.
- The agent connects to one address using MCP, the common standard for agent tools.
- Each customer picks which systems to connect. For private systems they install a small gateway that only makes outgoing connections.
- The customer sets a rule per action: allowed, needs approval, or blocked.
- The agent only sees the tools it was given. Sensitive calls wait for a person. Each request and decision is logged.
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
- An agent company's agent connects to Routant over MCP to reach a customer's private systems.
- The same agent calls Petrova for models and paid tools, on one balance.
- Routant's rules say what the agent may do in those systems. Petrova's holds and caps say what it may spend.
No partnership exists. We have not tested the two together. We could not try Routant, and Petrova is not live.
| Routant | Petrova | |
|---|---|---|
| Status | Website live. Sign-up by request. We could not test the product | No. Preview on test data |
| Reaches private systems inside a network | Yes, by its description | No |
| A person approves a single action | Yes, by its description | Planned, for paid tools only. Mock only |
| Each end customer sets their own rules | Yes, by its description | Sub-accounts planned. Mock only |
| Sells or routes model calls | No. It sells no models | Yes (test data) |
| Spend caps | Not found | Account, project, key, agent, session (test data) |
| Charge never above a hold set before the call | Not found | Yes (test data) |
| Approve by amount of money | Not found | Planned. Mock only |
| Public price | Enterprise: book a demo | List + 5.5% (not on sale) |
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 Routant is ahead
- Reaches private systems inside a network
- A person approves a single action
- Each end customer sets their own rules
What we could not check
- Whether its gateway, approvals and logs work. We did not sign up.
- Its customers and usage. None are named.
Read on 3 November 2026, on their public pages: routant.ai · routant.ai/about · routant.ai/terms · its Launch YC post. Things may have changed since. Each claim of theirs is their own statement.