Vanar, the AI infrastructure company, today announced the release of xBPP, a new open protocol built to govern autonomous AI agent behavior before transactions reach payment rails or APIs.
Open standard for pre‑execution policy control
Released under the Apache 2.0 license, the Execution Boundary Permission Protocol is available as an open standard that any developer, company, or platform can implement without permission from Vanar.
As AI systems take on a greater role in online transactions, the infrastructure that enables autonomous software to pay, access services, and complete actions is advancing quickly. The policy layer, which determines whether those actions should proceed in the first place, remains underdeveloped.
xBPP is designed to address the point. Vanar describes the protocol as a new category of infrastructure for the agent economy – Agent Policy Infrastructure. Existing systems can authorize transactions or execute them. Yet, they do not evaluate whether an autonomous system should be allowed to act before a request reaches payment rails or APIs.
The protocol grew out of work on autonomous payment systems. There, engineers found that current infrastructure could approve and carry out transactions, but lacked a standard way to enforce policy decisions ahead of execution.
Over the past year, major technology and payments companies have laid the groundwork for agent-driven commerce. Google’s AP2 provides a protocol for agent payments. The open x402 standard, governed by the Linux Foundation, and originally developed by Coinbase) enables HTTP-native payments between software and services. Stripe and OpenAI’s Agentic Commerce Protocol (ACP) supports agent-ready commerce flows. Mastercard’s Verifiable Intent introduces a trust layer for agentic commerce.
Together, these technologies enable agents to authorize and execute transactions. They do not determine whether a transaction should happen at all.
Before an enterprise gives an autonomous system authority to make financial decisions, it needs to know whether the action complies with spending policies, whether the counterparty is trusted, whether the request duplicates an earlier transaction, or whether human approval is required. xBPP is built to provide that decision point.
How xBPP evaluates agent actions
At its core, xBPP evaluates every proposed action against externally defined governance policies. The initial release includes a 1,760-line specification built around a structured nine-phase evaluation pipeline covering validation, risk checks, counterparty verification, escalation triggers, and final decision logic. The protocol also defines 67 standardized reason codes to explain outcomes and support auditable records.
Organizations manage policies outside the agent itself, which means rules can be updated without redeploying AI systems. When a transaction is proposed, xBPP returns one of three deterministic outcomes: ALLOW, BLOCK, or ESCALATE. This model gives enterprises a way to permit autonomous operation within defined boundaries. It also routes exceptional cases for human review. Each decision can be cryptographically signed and recorded to create a verifiable audit trail.
Following the release of xBPP, Jawad Ashraf, CEO of Vanar, said: “Many of the building blocks for agentic commerce are now in place, but policy enforcement has remained an open problem.”
“Before any business gives autonomous systems authority over payments or service access, it needs a reliable way to define boundaries and apply them consistently. It also needs a clear record of why a decision was made. xBPP was built to provide that control layer. It gives developers and enterprises a framework for enforcing decisions before execution, documenting outcomes, and expanding autonomy with the right safeguards in place.”
xBPP is designed to complement the emerging agentic commerce stack. In that stack, authorization protocols confirm user consent. Payment systems execute transactions. Audit systems record outcomes. xBPP sits between those layers as the policy decision point, determining whether an action should move forward before it reaches execution infrastructure.
The protocol is payment-rail agnostic and can operate across traditional banking networks, stablecoin systems, and emerging agent-native payment protocols.
A reference TypeScript SDK is available through the Vanar xBPP SDK repository, which also links to the npm package for @vanar/xbpp.
The complete xBPP specification is now publicly available at xbpp.org, along with developer tooling and reference implementations. Additional resources are also available in the VanarChain xBPP SDK repository.
