AI Practices AI实践 3d ago Updated 3d ago 更新于 3天前 48

Amazon Bedrock AgentCore payments is now generally available: Enabling agents to transact safely and autonomously at scale Amazon Bedrock AgentCore 支付功能现已正式推出:使代理能够安全、自主地大规模进行交易

Amazon Bedrock AgentCore payments is now generally available, enabling AI agents to autonomously transact for paid APIs, MCPs, and content with minimal integration effort Built in collaboration with Coinbase and Stripe, supporting both Coinbase and Stripe Privy stablecoin wallets for cost-effective microtransactions Expanded protocol support now includes the Machine Payment Protocol (MPP) alongside x402, plus a new "upto" spending ceiling scheme for dynamic pay-per-use pricing Infrastructure-lay Amazon Bedrock AgentCore payments正式发布,使AI agent能够自主完成微额支付交易 集成Coinbase和Stripe Privy钱包,支持USDC稳定币和传统支付方式为agent充值 新增支持Machine Payment Protocol (MPP)和x402的"upto"动态定价方案 支付会话机制提供预算上限和过期时间双重保护,防止agent非确定性行为导致超支 通过MCP server和AgentCore Gateway实现付费API端点的可发现性

72
Hot 热度
62
Quality 质量
68
Impact 影响力

Analysis 深度分析

TL;DR

  • Amazon Bedrock AgentCore payments is now generally available, enabling AI agents to autonomously transact for paid APIs, MCPs, and content with minimal integration effort
  • Built in collaboration with Coinbase and Stripe, supporting both Coinbase and Stripe Privy stablecoin wallets for cost-effective microtransactions
  • Expanded protocol support now includes the Machine Payment Protocol (MPP) alongside x402, plus a new "upto" spending ceiling scheme for dynamic pay-per-use pricing
  • Infrastructure-layer payment sessions enforce deterministic spend caps and expiry times, addressing the non-deterministic nature of agent behavior
  • Integrated observability via CloudWatch and AgentCore Observability provides audit trails, transaction health dashboards, and spending visibility for production workloads

Why It Matters

This addresses a critical gap in the agentic AI ecosystem: while agents have become sophisticated at reasoning and tool selection, they lacked a secure, scalable mechanism for autonomous payment. As AI agents evolve into long-running, autonomous systems and the industry shifts toward pay-per-use microtransaction models, AgentCore payments provides the financial infrastructure layer that enables agents to access paid content, APIs, and services without human intervention—unlocking new business models for content providers and expanding the practical utility of agentic systems.

Technical Details

  • Wallet Integration: Supports Coinbase and Stripe Privy stablecoin wallets; developers provide API keys/secrets stored in AgentCore Identity Secrets Manager, with agents receiving only short-lived tokens for transaction signing—raw credentials are never exposed to the agent
  • Protocol Support: Protocol-agnostic architecture; at GA supports x402 (including the new "upto" scheme for spending ceilings and dynamic pricing) and the Machine Payment Protocol (MPP) co-authored by Stripe and Tempo, allowing integration once for multiple merchant protocols
  • Payment Sessions: Each transaction runs within a scoped payment session with two configurable caps—a maximum spend amount and an expiry time—with deterministic budget checks at the infrastructure layer before any payment signing occurs
  • Discoverability: Pay-per-use x402 endpoints are exposed as an MCP server via AgentCore Gateway, now enhanced with a curated list ranked by social proof, metadata richness, description quality, and availability
  • Observability: Automatically emits logs to Amazon CloudWatch and spans to AgentCore Observability, with prebuilt dashboards tracking transaction success rates, average transaction values, and spending across agents, sessions, and time periods

Industry Insight

  • The shift from human-centric subscription pricing to machine-to-machine microtransactions will create new revenue streams for content and API providers; enterprises building agent-based solutions should prioritize integrating agentic payment capabilities early to access paywalled data and premium APIs
  • The non-deterministic nature of LLM agents makes infrastructure-layer spending controls essential—any production agent system handling payments must implement deterministic caps and audit trails to prevent runaway costs from retries or misinterpretations
  • Protocol standardization (x402, MPP) around machine payments is consolidating, and developers who adopt protocol-agnostic payment abstractions will be better positioned to integrate across the growing ecosystem of pay-per-use AI services and content platforms

TL;DR

  • Amazon Bedrock AgentCore payments正式发布,使AI agent能够自主完成微额支付交易
  • 集成Coinbase和Stripe Privy钱包,支持USDC稳定币和传统支付方式为agent充值
  • 新增支持Machine Payment Protocol (MPP)和x402的"upto"动态定价方案
  • 支付会话机制提供预算上限和过期时间双重保护,防止agent非确定性行为导致超支
  • 通过MCP server和AgentCore Gateway实现付费API端点的可发现性

为什么值得看

AI agent从工具调用向自主交易演进的关键基础设施突破,解决了agent经济闭环的最后一环。对开发者而言,这意味着可以用几行代码为agent添加支付能力,开启真正的agent-to-agent经济模式。

技术解析

  • 钱包集成架构:通过AgentCore Identity Secrets Manager存储开发者凭证,agent不直接接触原始密钥,使用短期令牌驱动Coinbase/Stripe Privy钱包执行交易签名操作。
  • 协议支持扩展:GA版本新增MPP(Machine Payment Protocol)支持,同时x402协议新增"upto"方案,允许agent设置消费上限而非固定价格,实现真正的按推理计费。
  • 支付会话机制:每次agent交互在独立的支付会话中执行,会话配置最大金额和过期时间两个硬性上限,在基础设施层进行确定性预算检查。
  • 可观测性集成:自动将支付日志发送至CloudWatch,交易span发送至AgentCore Observability,提供预置仪表板展示交易成功率、平均交易价值等关键指标。
  • 端点发现系统:通过AgentCore Gateway将Coinbase的x402付费端点以MCP server形式暴露,基于社会证明、元数据丰富度等维度筛选高质量端点。

行业启示

  • Agent经济基础设施正在成熟,支付能力将成为agent平台的核心竞争力,早期布局者将建立生态壁垒。
  • 微额支付(cents级)和动态定价模式将催生新的内容和服务商业模式, publishers可通过agent支付解锁付费内容。
  • 安全与可控性设计(凭证隔离、预算上限、会话隔离)是agent支付规模化落地的关键前提,企业级采用需重点关注这些机制。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Agent Agent Product Launch 产品发布 Security 安全 LLM 大模型