Capri Ai Agentpay

Capri Ai Agentpay is a governed payment layer purpose built for the emerging ecosystem of autonomous AI agents. As artificial intelligence systems increasingly operate outside the boundaries of simple prompts and begin to call paid APIs, settle transactions, and interact with live infrastructure, the question of financial control becomes paramount. Handing a private key to an agent is a security vulnerability waiting to be exploited; a single malicious or poorly constructed prompt can drain an entire wallet in seconds. Capri Agentpay solves this by inserting a sophisticated runtime between the agent and the signing authority. Every payment request is intercepted before any signature is created, evaluated against a predetermined policy framework, and only settled when all conditions are met. The product is designed for engineering teams and platform operators who are shipping agent systems in production and require real financial controls, audit trails, and human oversight. It supports HTTP 402 Payment Required challenge handling, per agent budgets and spending caps, merchant allowlists, human in the loop approval workflows, and verifiable onchain receipts. Agents never hold keys, and every transaction is policy checked before execution. Capri Agentpay is currently shipping and available in a Base Sepolia sandbox environment for testing and integration.

x402 Payment Challenge Handling

Capri Agentpay natively handles the HTTP 402 Payment Required status code, which is emerging as the standard protocol for machine to machine payments. When an agent makes a request to a paid API or MCP tool and receives a 402 challenge, the runtime intercepts the response, evaluates the applicable policy, and if approved, settles the required payment. The original request is then automatically retried with cryptographic proof of payment attached. This eliminates the need for developers to write custom payment logic for every paid endpoint and ensures a seamless, standards compliant experience for agent interactions.

Per Agent Budgets and Spending Caps

Operators can define granular financial boundaries for each individual agent or agent group. These budgets and caps are enforced at the runtime level, meaning no agent can exceed its allocated spending limit regardless of the requests it makes. The policy evaluation follows a fixed order: merchant allowlist, per transaction cap, daily budget, and finally an auto approve threshold. This layered approach ensures that even if an agent is compromised or behaves unexpectedly, financial exposure is strictly contained within the defined parameters. Budgets can be configured and adjusted through the operator interface or programmatically via the API.

Merchant Allowlists and Human in the Loop Approvals

Capri Agentpay provides a merchant allowlist that restricts which endpoints or services an agent can pay. Any request to a merchant not on the allowlist is automatically denied. For transactions that exceed the auto approve threshold, the runtime pauses execution and routes the request to a human operator for review. The operator sees the exact amount, asset, chain, and recipient before making a decision, along with the policy branch that triggered the review. This hybrid approach combines the speed of automated approvals for low risk payments with the safety of human judgment for higher value or suspicious transactions.

Verifiable Receipts and Six MCP Tools

Every settled payment generates a signed receipt that contains hashes, timestamps, and verification fields. These receipts can be verified offline, providing a cryptographic audit trail for every transaction. This is critical for compliance, accounting, and security audits. Instead of exposing raw key access to agents, Capri provides six narrow MCP (Model Context Protocol) tools that agents can call. These tools are specifically designed for payment related operations and prevent agents from performing any action outside their intended scope. Additionally, Capri offers a VS Code extension for developers to integrate and test payment controls directly within their development environment.

Autonomous Research Agents Accessing Premium Data

A research agent operating in a financial or scientific domain needs to call premium data APIs that charge per request. With Capri Agentpay, the agent can autonomously fetch data while the runtime enforces a daily budget of fifty dollars and an auto approve threshold of five dollars per transaction. If the agent attempts to access a new, expensive dataset, the request is paused for human review. The operator sees the specific API endpoint, the cost, and the agent identity before approving or denying the transaction.

Multi Agent Workflows with Shared Budgets

In a deployment where multiple agents collaborate on a complex task, such as supply chain optimization or content generation, each agent may need to call different paid services. Capri Agentpay allows operators to define a shared daily budget for the entire agent team while setting individual per transaction caps for each agent. This prevents any single agent from consuming the entire budget and ensures that spending is distributed according to the workflow requirements. The runtime tracks all payments and provides a consolidated view of spending across the agent fleet.

Internal Tool Automation with Compliance Controls

Enterprise teams often deploy agents to automate interactions with internal tools and SaaS platforms that have usage based pricing. Capri Agentpay provides the compliance and security controls required for production deployments. Every payment is policy checked, logged, and tied to a verifiable receipt. The human in the loop approval feature ensures that any unusual spending pattern or high value transaction is reviewed by a responsible operator before funds are released.

Developer Sandbox Testing of Agent Payment Flows

Developers building agent systems can use the Base Sepolia sandbox environment to test payment flows, policy configurations, and approval workflows without risking real funds. The sandbox provides the same runtime, policy engine, and receipt generation as the production environment. This allows teams to iterate on their agent payment logic, validate policy rules, and train operators on the review interface before deploying to mainnet.

What happens if an agent tries to make a payment that violates the policy?

If a payment request violates any policy check, such as exceeding the daily budget, hitting a merchant not on the allowlist, or surpassing the per transaction cap, Capri Agentpay automatically denies the transaction. The agent receives a denial response, and the event is logged in the audit trail with details about which policy check failed. No signature is created, and no funds are moved. The operator can review the denied request in the interface and adjust policies if needed.

How does the human in the loop approval process work?

When an agent initiates a payment that exceeds the configured auto approve threshold, the runtime pauses execution and creates a review request. The operator receives a notification and can open the review queue in the Capri interface. The operator sees the exact amount, asset, chain, and recipient, along with the policy branch that triggered the review. They can approve or deny the request with a single click. If approved, the payment is settled and the agent's original request is retried. If denied, the agent receives a denial response.

Can I use Capri Agentpay with any blockchain or payment network?

Capri Agentpay is currently shipping with support for Base Sepolia as the sandbox settlement environment. The architecture is designed to support multiple chains and payment networks over time. The x402 protocol and the policy engine are chain agnostic, meaning the same runtime can be configured to settle payments on different networks as support is added. The verifiable receipts are chain aware and include the specific chain and asset information for each transaction.

What is the difference between Capri Agentpay and giving an agent a private key?

Giving an agent a private key grants it unlimited and unsupervised access to the associated wallet. A single bad prompt or compromised agent can drain the entire balance. Capri Agentpay eliminates this risk by placing a governed runtime between the agent and the signing authority. The agent never holds a key. Every payment request is intercepted, policy checked, and only signed if all conditions are met. Additionally, Capri provides audit trails, human oversight, and narrow MCP tools that restrict the agent to only payment related actions.

Capri Ai Agentpay is currently available by request. The product is opening access in small waves to teams building agent systems that need payment controls, approval flows, or operator facing runtime boundaries. Specific pricing plans, tiers, and costs are not publicly listed at this time. Interested teams can request access through the Capri AI website, book a demo, or install the VS Code extension to begin exploring the sandbox environment.