PayGraph

PayGraph is an open-source SDK designed to provide policy-controlled spending, approvals, and audit logs for AI agents that interact with real-world financial transactions. It addresses the need for secure and auditable purchasing power when deploying autonomous agents, ensuring that every transaction is governed by explicit policies and logged for review. The tool is particularly aimed at AI engineers, startups, and teams building agent workflows that require strict controls over how agents can spend money.

The SDK features a policy engine that allows developers to set spending limits, designate approved merchants, and impose budget caps. Each transaction initiated by an agent is evaluated against these policies before being processed, with every attempt—whether approved or denied—recorded in an audit trail. This audit log captures which agent made the request, which policy was checked, the result of the decision, and the reasoning behind it. The control logic operates outside the context of the language model, providing an added layer of security against manipulation, such as prompt injection or malicious tool responses.

PayGraph supports a pluggable payment gateway interface and ships with integration for Stripe Issuing. Developers can bring their own gateway, and support for MPP (Multi-Party Payments) is noted as being on the roadmap. The SDK is distributed via pip for Python, and includes examples for integrating with agent frameworks such as LangGraph. The tool is MIT-licensed, encouraging open-source contributions and community involvement.

By serving as a control plane rather than a simple payment wrapper, PayGraph enables teams to enforce hard spending rules and maintain full auditability before deploying card issuance or financial operations in production environments.