Capri Ai Agentpay

Capri AgentPay is the first governed payment layer built specifically for AI agents. When your agent needs to call a paid API or settle a transaction on its own, giving it a private key is a security nightmare. One bad prompt or a hallucinated command could drain your wallet in seconds. Capri AgentPay solves this by placing a secure runtime between your agent and the signer. Instead of giving your agent raw key access, you give it a typed interface. When the agent hits an HTTP 402 Payment Required challenge, Capri automatically intercepts it. It then evaluates your predefined policy in a fixed order: first checking a merchant allowlist, then a per-transaction cap, then a daily budget, and finally an auto-approve threshold that can pause for human review. Only when every single check passes does Capri settle the payment, retry the original request with proof attached, and store a signed receipt you can verify offline. This product is designed for engineering teams building agent systems that need to interact with paid APIs, internal tools, and stateful systems without exposing their infrastructure to risk. Capri AgentPay gives you operator-grade controls like per-agent budgets, merchant allowlists, human-in-the-loop approvals, and verifiable receipts. It ships with six MCP tools, a VS Code extension, and a Base Sepolia sandbox for testing. With Capri, your agents never hold your keys. Every payment is policy-checked before a signature ever exists. This is the safe, scalable way to let your agents spend money.

x402 Payment Handling

Capri AgentPay natively handles the HTTP 402 Payment Required status code. When your agent makes a request to a paid API and receives a 402 challenge, Capri automatically intercepts it. It evaluates your policy gate, settles the payment using the agreed asset and chain, and then retries the original request with cryptographic proof of payment attached. This means your agent can seamlessly access premium data or services without you having to manually approve every single transaction. The entire flow happens in the background, guided by your rules.

Per-Agent Budgets and Caps

You can set individual budgets and per-transaction spending caps for each of your agents. This feature gives you granular control over how much each agent can spend in a given day or on a single API call. For example, you can give your research agent a daily budget of $50 with a per-call cap of $5, while your data analysis agent might have a higher limit. These budgets are enforced automatically by the policy runtime, preventing any single agent from overspending or making a costly mistake.

Merchant Allowlists

Capri AgentPay includes a merchant allowlist feature that lets you specify exactly which APIs or services your agents are allowed to pay. You create a list of approved merchant addresses or domains. If an agent tries to pay a merchant that is not on the allowlist, the transaction is automatically denied. This is a critical security layer that prevents your agent from sending funds to unknown or malicious endpoints. It is the first check in the policy evaluation order, ensuring only trusted recipients ever receive a payment.

Human-in-the-Loop Approvals

For transactions that exceed your auto-approve threshold, Capri AgentPay pauses and requests human review. The operator sees the exact amount, asset, chain, and recipient before making a decision. They can approve, deny, or adjust the request. This feature provides a safety net for high-value or unexpected payments. It combines the speed of automation for routine transactions with the oversight of a human for anything that falls outside your predefined comfort zone. This is essential for maintaining control in production environments.

Autonomous Research Agents Accessing Premium Data

Imagine you have an AI agent that needs to query premium financial data APIs to generate market reports. With Capri AgentPay, you can give this agent a daily budget of $100 and a per-call cap of $10. You add the financial data API to your merchant allowlist. Every time the agent hits a 402 challenge, Capri checks the policy, approves the payment if within limits, and retries the request. The agent gets the data, you get a signed receipt, and your wallet is never directly exposed. This lets you scale your research operations safely.

Internal Tool Automation with Spending Controls

Your engineering team might have an agent that automates deployments or runs test suites on paid cloud services. Instead of giving the agent a set of API keys with broad permissions, you use Capri AgentPay. You set a strict daily budget and a per-transaction cap. The agent can only pay for services on your internal allowlist. If a new, unapproved service is called, the transaction is blocked. This prevents accidental or malicious spending on external resources and keeps your cloud costs predictable and under control.

Customer Support Agents Resolving Issues with Paid Services

A customer support agent might need to temporarily access a paid data enrichment service to look up a customer's account details. With Capri AgentPay, you can configure a policy that allows small, frequent transactions to a specific enrichment API. The agent can handle the request autonomously. If the cost exceeds a small threshold, the system pauses and asks a human supervisor to approve the payment. This blends AI efficiency with human oversight, ensuring customer issues are resolved quickly without compromising financial controls.

Multi-Agent Systems with Shared Budgets

In a complex system, you might have multiple agents working together on a single project, like analyzing a dataset from multiple paid sources. Capri AgentPay allows you to assign a shared budget to a team of agents. Each agent has its own per-call cap, but they all draw from the same daily pool. The policy engine tracks spending across all agents in real time. If one agent tries to spend the remaining budget on a single call that exceeds its cap, the transaction is denied. This prevents any single agent from monopolizing the team's resources.

How does Capri AgentPay protect my private keys?

Capri AgentPay places a governed runtime between your agent and the signer. Your agent never has direct access to your private key. Instead, it calls a typed interface. When a payment is needed, the runtime evaluates your policy. Only after every check passes does it create a signature. This means a bad prompt or a compromised agent cannot drain your wallet because the key is never exposed to the agent's logic.

What happens when a transaction exceeds my auto-approve threshold?

When a transaction exceeds your auto-approve threshold, Capri AgentPay pauses the request and sends it to a human review queue. The operator sees the exact amount, asset, chain, and recipient. They can approve, deny, or adjust the payment. Once approved, the system settles the payment and retries the original request. This provides a manual safety net for high-value or unexpected transactions while keeping routine payments fast and automated.

Can I test Capri AgentPay without using real money?

Yes. Capri AgentPay includes a Base Sepolia sandbox for settlement. You can deploy your agents and test your policies in a testnet environment using test tokens. This allows you to validate your budgets, caps, allowlists, and approval workflows before moving to a live network. The sandbox gives you full confidence that your system works correctly without any financial risk.

What kind of receipts does Capri AgentPay provide?

Capri AgentPay stores a signed receipt for every settled payment. This receipt includes hashes, timestamps, and verification fields. You can download and verify these receipts offline to prove that a specific payment was made to a specific merchant under a specific policy. This is invaluable for audit trails, compliance, and reconciling your spending. The receipts are cryptographically signed, so they cannot be forged.