Capri Ai Agentpay

Capri Ai Agentpay is a governed payment layer specifically designed for AI agents. As autonomous agents increasingly interact with paid APIs, internal tools, and stateful systems, the traditional approach of handing them a private key creates a critical security risk. A single bad prompt or compromised instruction can drain an entire wallet with no recourse. Capri Agentpay solves this fundamental problem by placing a runtime layer between the agent and the signing mechanism. This runtime intercepts every payment request, evaluates it against a fixed set of policies, and only settles transactions that pass all checks.

The product is built for teams shipping agent systems in production environments where money, approval workflows, and live infrastructure are involved. It handles HTTP 402 Payment Required challenges automatically, allowing agents to call paid APIs without ever holding raw keys. The system follows a strict policy evaluation order: merchant allowlist, per-transaction cap, daily budget, and then an auto-approve threshold that can pause for human review. Only when every check passes does Capri settle the payment, retry the original request with proof attached, and store a signed receipt that can be verified offline.

Capri Agentpay provides six MCP tools instead of exposing raw key access, giving developers narrow, typed surfaces for payment operations. It also includes a VS Code extension for developer convenience and supports Base Sepolia sandbox settlement for testing. The product is designed for operators who need clear review queues, audit trails, and controls that do not collapse into manual busywork. It is the first shipping layer in the Capri AI stack, addressing the messy parts that show up the moment an agent touches money or approval workflows.

Policy-Governed Payment Runtime

Capri Agentpay places a runtime between the agent and the signer, evaluating every payment request against a fixed policy order. The system checks the merchant allowlist first, then the per-transaction cap, followed by the daily budget, and finally the auto-approve threshold. If the request exceeds the threshold, it pauses for human review. Only when all checks pass does the system create a signature and settle the payment. This ensures that no transaction occurs without explicit policy approval, eliminating the risk of a single bad prompt draining funds.

HTTP 402 Payment Required Handling

The product automatically intercepts HTTP 402 Payment Required challenges from paid APIs and MCP tools. When an agent calls a typed surface and encounters a 402 response, Capri evaluates the payment request against its policies, settles the transaction if approved, and retries the original request with proof of payment attached. This seamless handling allows agents to access premium data and services without developers needing to manually manage payment flows or expose sensitive credentials.

Per-Agent Budgets and Spending Caps

Capri Agentpay allows operators to set individual budgets and spending caps for each agent. These per-agent limits include a daily budget that resets and a per-transaction cap that prevents any single payment from exceeding a defined amount. Operators can configure these limits through the policy system, ensuring that each agent operates within its allocated financial boundaries. This granular control prevents runaway spending and provides clear accountability for agent-driven transactions.

Human-in-the-Loop Approval Workflow

When a payment request exceeds the auto-approve threshold, Capri pauses the transaction and sends it to a human operator for review. The operator sees the exact amount, asset, chain, and recipient before making a decision. They also see which policy branch passed or failed the request. This human-in-the-loop workflow provides an essential safety net for high-value or unusual transactions, allowing teams to maintain oversight without blocking routine payments that fall within policy boundaries.

Verifiable Signed Receipts

Every settled transaction generates a signed receipt that can be verified offline. The receipt object contains hashes, timestamps, and verification fields that allow operators to independently confirm that a payment was authorized and completed. This feature is critical for audit trails, compliance requirements, and security reviews. It gives teams the ability to prove that every payment went through the proper policy gate, providing transparency and trust in agent-driven financial operations.

Six MCP Tools for Payment Operations

Instead of exposing raw key access, Capri Agentpay provides six narrow MCP (Model Context Protocol) tools that cover all necessary payment operations. These tools include functions for fetching data with payment handling, checking balances, reviewing pending approvals, and inspecting receipts. This design ensures that agents have exactly the capabilities they need to make payments while preventing any direct access to signing keys or wallet controls. The tools are typed and documented, making them easy to integrate into existing agent systems.

Autonomous Research Agents Accessing Paid APIs

Research agents that need to query premium data sources or paid APIs can use Capri Agentpay to handle payments automatically. The agent calls the API through the Capri runtime, which intercepts any 402 Payment Required challenges, evaluates the request against policy, and settles the payment if approved. This allows research teams to deploy agents that can access expensive datasets without manual intervention, while maintaining strict budget controls and audit trails for every transaction.

Internal Tool Automation with Spending Limits

Teams building agents that interact with internal tools and paid services can use Capri Agentpay to set per-agent budgets and caps. For example, an agent that generates reports using a paid image generation API can have a daily budget of $50 and a per-transaction cap of $5. If the agent tries to exceed these limits, the payment is automatically denied. This use case is ideal for organizations that want to automate workflows while keeping spending predictable and under control.

Customer-Facing Agent Systems with Payment Flows

Companies deploying customer-facing agents that need to make payments on behalf of users can use Capri Agentpay to enforce policies and provide human oversight. For instance, a travel booking agent that purchases tickets or hotel rooms can be configured with an auto-approve threshold of $200. Transactions above that amount trigger a human review, allowing a customer service representative to approve or deny the payment. This balances automation with the need for human judgment in high-value transactions.

Developer Sandbox Testing for Agent Payments

Development teams building and testing agent systems that handle payments can use the Base Sepolia sandbox settlement feature of Capri Agentpay. This sandbox environment allows developers to simulate payment flows, test policy configurations, and verify receipt generation without using real funds. Teams can experiment with different budget settings, approval thresholds, and merchant allowlists before deploying to production. This reduces the risk of costly mistakes during development and ensures that payment policies work as intended.

How does Capri Agentpay protect against a bad prompt draining a wallet?

Capri Agentpay places a runtime between the agent and the signer. The agent never holds a private key or has direct access to signing capabilities. Every payment request must pass through the policy gate, which checks the merchant allowlist, per-transaction cap, daily budget, and auto-approve threshold in a fixed order. Only when all checks pass does the system create a signature and settle the payment. This means that even if a bad prompt instructs the agent to make a payment, the policy system will block it if it violates any rule.

What happens when a payment exceeds the auto-approve threshold?

When a payment request exceeds the auto-approve threshold, Capri pauses the transaction and sends it to a human operator for review. The operator sees the exact amount, asset, chain, and recipient before making a decision. They also see which policy branch passed or failed the request. The operator can then approve, deny, or modify the transaction. This human-in-the-loop workflow ensures that high-value or unusual payments always receive manual oversight before being settled.

Can I verify a payment receipt offline?

Yes, every settled transaction generates a signed receipt that can be verified offline. The receipt contains hashes, timestamps, and verification fields that allow you to independently confirm that a payment was authorized and completed through the proper policy gate. This is useful for audit trails, compliance requirements, and security reviews. You do not need to be connected to the Capri system to verify the authenticity of a receipt.

What is the Base Sepolia sandbox and how do I use it?

The Base Sepolia sandbox is a test network settlement option provided by Capri Agentpay. It allows developers to simulate payment flows, test policy configurations, and verify receipt generation without using real funds or interacting with mainnet. You can configure your agent to use the Base Sepolia sandbox during development and testing. Once you are satisfied with your policy setup and payment flows, you can switch to mainnet settlement for production use. This sandbox environment reduces the risk of costly mistakes during development.