Introducing Conto: Controls for AI Agents that Pay

Agents are doing more than ever before at home and at work. Now they're starting to pay for things on their users' behalf. That raises a few questions. How much can the agent spend? Who can it pay? When should it ask first? And where did the money go?

Today we're launching Conto to enable organizations to build and manage agents that pay. You set the budget, who the agent can pay, and when it needs approval. Conto checks each payment request against those rules before the agent's payment tools can transact, and keeps a record of what happened.

Agents Are Taking Action

Until recently, agents could tell you what to buy, but they couldn't buy it. Better browser use is changing that and kickstarting agentic commerce. Agents can now use cards to check out on the websites people already use while stablecoins and protocols like x402 and MPP let them pay for digital services as they go.

Personal agents like Instinct and Muse are the first movers in agentic payments. Once an assistant can reorder groceries or book a trip on its own, people start asking why their company still handles routine purchases manually or by workflow. Adoption of agentic payments will move in stages, from digital services to everyday purchases to business workflows where the largest volumes are.

The ability to pay is arriving faster than the controls around it, and that gap is where things go wrong.

Dealing With Money Requires Trust

Right now, most people have two choices. They can approve every payment, which makes the experience clunky and not much better than browsing manually. Or they can give it a card and hope things go right, which few people will do with real money and most companies won't allow. Neither lets anyone hand over a whole task.

The future will be a bit of both: agents working on their own inside the guardrails you set, with people stepping in only for exceptions. That needs real controls.

Conto: The Control Center for Agentic Payments

Conto gives developers and teams building agents the controls to make them trustworthy when it comes to payments. With Conto, an agent checks each request before it pays. Conto evaluates it against your rules and policies, approves it, blocks it, or sends it for human review, and logs all the transaction details. This is powered by the Conto platform which includes:

• Policy engine: fine grained spend policies, budgets, per-purchase limits, approved recipients, and approval thresholds for each agent. See the policy engine.
• Multiple rails: pay with a card you already have, without the card number ever reaching the agent, or use stablecoins for low-cost, high-frequency payments and x402/MPP micropayments. See connected cards.
• Network intelligence: see who an agent is paying and set different rules for new and familiar recipients. See network intelligence.
• Audit and governance: every request, decision, and approval in one place. See audit and reconciliation.

Launch offer: a free month on the Builder plan, which includes connected cards. Contact us to get started.

What Makes Conto Different

Most spending tools were built for people. A company card has a limit, and an expense report shows up after the money is gone. Conto was built for agents. It looks at each payment before it happens, knowing which agent is asking and what it's trying to do.

We also look at who's getting paid, not just how much. Conto scores each recipient using your own payment history and anonymized signals from across the Conto network, so an unfamiliar or risky payee gets a second look before any money moves.

And you don't need a separate setup for every way an agent pays. Cards you already have, stablecoins, and pay-per-call protocols like x402 and MPP are managed in one place and show up in the same record. There are many different rails on which agents can transact, and we want to build the controls to manage all of them.

Developers connect agents through the SDK, API, MCP server, or skills.

So What Can You Build With Conto?

From personal agents to agents at work in the enterprise, the pattern is the same: the agent does the work, and your rules decide what it can pay for.

Personal Agents That Shop Within Your Limits

For teams building commerce agents, Conto gives each of your users their own controls. They set the limits, and your agent works inside them. That makes "buy it for me" something users can say yes to. See spend controls for personal assistants.

A shopping assistant can reorder household basics every week within a set budget, only from stores you've approved, and ask before anything over $100. A travel agent can book a flight under your fare limit on its own, then wait for your yes before a nonrefundable hotel. With a connected card, it checks out on the merchant's normal site without your card number ever reaching the agent. Try the shopping demo or plan a trip with Claude.

Agents at Work

Agents for work will do more and more as they move to parity with human employees. Inside a company, agents will take on payment work that runs every day. Each kind of agent needs its own rules.

Treasury: A treasury agent keeps the company's wallets funded for the day's payments and moves money between them as balances change. It can only send to the company's own approved wallets, only during business hours, and anything above a daily limit waits for the treasury team. Try the treasury demo.

Marketplace payouts: A marketplace agent pays sellers, drivers, or creators as their work is completed, and issues refunds and credits when orders go wrong. It pays only recipients on its approved list, caps each payout, and holds the first payout to a new recipient for review. Limits on how often it can pay stop a runaway loop before it drains the account. Try the payouts demo.

Procurement and payables: A procurement agent reorders supplies from approved vendors within a monthly budget and pays their invoices on schedule. Orders from a new vendor, and any change to a vendor's payment details, wait for review, since a swapped account on a fake invoice is a common fraud. Try the procurement demo or see vendor controls.

Explore more use cases, or try the demos with your own spending rules.

Trust Your Agents to Transact

Agents are paying for things, first for personal errands and next for the everyday work of running a company. How much people hand over will come down to trust, and trust comes from rules they can set, check, and change. If you're interested in how Conto can help you build trusted agents that pay, get in touch!

Get Started

Try Conto for yourself:

1. See a policy in action. Run a payment through the interactive demo and see how your rules impact it.
2. Make a test payment. Build with Conto walks you through connecting an agent and setting its spending rules in a sandbox.
3. Set up your team. Create a free account to add your agents and set their budgets and approvals. For connected cards, a free month on Builder, or a company-wide rollout, get in touch.

Your agents pay. Your rules apply.

Related Reading

• The Era of Agents that Take Action: how computer use brings agents to checkout.
• The Enterprise is the Biggest Opportunity for Agentic Payments: why company workflows are the largest market for agents that pay.
• Spend Management for AI Agents: the rules to set before an agent spends.
• Claude commerce demos: shopping, travel, and ticketing agents running with Conto controls.