EILMELDUNGEN LIVE
🕵️ Sicherheitslücken[webapps] Langflow 1.8.4 - Path Traversal to Remote Code Execution(31.08.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] C-MOR 6.0104 - Cross-Site Scripting (XSS)(31.08.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] CubeCart 6.7.4 - Stored XSS(31.08.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] CubeCart 6.7.4 - Cross-Site Scripting(31.08.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] Wolf CMS 0.8.3.1 - RCE v(01.09.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] Bludit CMS - Stored XSS(01.09.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] Grav CMS 2.0.7 - RCE(01.09.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] miniOrange 5.4.3 - Unauthenticated Auth Bypass(01.09.2026 um 02:00 Uhr)
🕵️ SicherheitslückenPaperCut vulnerability poc.(30.08.2026 um 19:29 Uhr)
🕵️ Reverse EngineeringReverse Engineering the Auto-Color Linux Backdoor(02.09.2026 um 07:00 Uhr)
🕵️ Sicherheitslücken[webapps] Langflow 1.8.4 - Path Traversal to Remote Code Execution(31.08.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] C-MOR 6.0104 - Cross-Site Scripting (XSS)(31.08.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] CubeCart 6.7.4 - Stored XSS(31.08.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] CubeCart 6.7.4 - Cross-Site Scripting(31.08.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] Wolf CMS 0.8.3.1 - RCE v(01.09.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] Bludit CMS - Stored XSS(01.09.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] Grav CMS 2.0.7 - RCE(01.09.2026 um 02:00 Uhr)
🕵️ Sicherheitslücken[webapps] miniOrange 5.4.3 - Unauthenticated Auth Bypass(01.09.2026 um 02:00 Uhr)
🕵️ SicherheitslückenPaperCut vulnerability poc.(30.08.2026 um 19:29 Uhr)
🕵️ Reverse EngineeringReverse Engineering the Auto-Color Linux Backdoor(02.09.2026 um 07:00 Uhr)

29 🕛 kürzlich 11 Min Lesezeit CVE-RADAR
0

How t54 built a trust layer with Amazon Bedrock AgentCore payments

↗ Quelle (Artificial Intelligence)
🗣️ Stimme:
📑 Inhaltsübersicht

An agentic system can research, reason, and orchestrate multi-step workflows, but the moment it hits a paywall, it stops. It has no wallet, no card, and no spending limit. t54 solved that problem. without a human approving a single one. Each transaction is a micropayment between $0.001 and $0.01: fast, high-volume calls that no human could review at that rate.


The problem: Giving an agent a wallet is straightforward, but governing it at scale is not


t54’s customers deploy agentic systems that need to pay for third-party services autonomously. Consider a financial services team running an agentic system that monitors stock portfolios and alerts analysts when positions shift. The system needs real-time market data from a paid API. Giving the agent access to a wallet is straightforward. The hard part is everything around it: enforcing spending limits so a single misconfigured loop cannot drain an account, isolating credentials so the agent never touches raw keys, auditing every transaction for compliance, and doing all of this across dozens of agents calling hundreds of endpoints.


Without that governance layer, the team faces a choice: review and approve every transaction manually, or accept unmanaged risk. At low volumes, manual approval works. At machine speed, with thousands of API calls per hour, it breaks down. The team needs autonomous payments with built-in controls.


t54 builds trust infrastructure for agent payments. Their customers kept asking for the same thing: a way to let agents pay that was secure, scalable, and provider-agnostic, without building custom payment orchestration from scratch. To deliver that, t54 needed a payments backbone with session-scoped spending limits, credential vaulting, and multi-provider support baked in.


Amazon Bedrock AgentCore payments provides the spending infrastructure: session limits, credential isolation, and payment execution. t54’s x402-secure provides the trust intelligence: real-time endpoint scoring that determines who is safe to pay. Together, they give teams governed agent payments without custom integration work.


What t54 built



Figure 1: Trust-gated x402 payment flow, where the trust check must pass before Amazon Bedrock AgentCore payments settles a transaction



This diagram shows the end-to-end flow across five lanes: the application backend, Amazon Bedrock AgentCore runtime (a capability of Amazon Bedrock AgentCore), the Strands agent with its deterministic trust gate, external paid services, and Amazon Bedrock AgentCore payments. A trust check must pass before a payment can proceed. For the full Amazon Bedrock AgentCore payments architecture, including AWS Identity and Access Management (IAM) authentication, AWS Secrets Manager credential storage, and Amazon CloudWatch observability, see . The deterministic trust gate, IAM role separation, and session spending limits keep every transaction governed without requiring a human in the loop.



“Twenty million transactions tells us agent-to-agent payments are already happening at real scale, not just in a demo. Each one is a micropayment between $0.001 and $0.01, the kind of fast, small call for data or an API that no person could review in real time. Along the way we’ve blocked payments to endpoints that scored as high-risk, so the session keeps its spending limit and the agent gets routed somewhere safer.”



— Chandler Fang, founder of t54


The financial primitives for autonomous AI are already in production. The harder question is not whether agents will need to pay for things. It is whether the governance layer is in place before something settles that should not.


Under the hood: Control plane and data plane


For builders who want to implement a similar integration, here is how t54 mapped their system to the two layers of Amazon Bedrock AgentCore payments.


Setting up (control plane)


Three resources create payment infrastructure without custom integration code per provider:



  • Credential Provider stores credentials in a token vault. Credentials do not appear in plain text to the agent runtime.

  • Payment Manager ties together authorization, identity, and payment connectors. t54 configured it with a CUSTOM_JWT authorizer backed by an OpenID Connect (OIDC) discovery endpoint.

  • Payment Connector specifies the payment provider type (CoinbaseCDP) and references the Credential Provider, linking the Payment Manager to the external provider.


At runtime (data plane)


Three operations handle live transactions:



  • CreatePaymentSession opens a session with a spending limit, an expiry window (15 to 480 minutes), and a userId. Amazon Bedrock AgentCore payments tracks the available spend in real time.

  • CreatePaymentInstrument provisions an Embedded Crypto Wallet with a specified network. The response returns a wallet address and a redirect URL for onboarding.

  • ProcessPayment executes the payment. Before it settles, Trustline scores the endpoint. When approved, Amazon Bedrock AgentCore payments returns a processPaymentId, status, and full audit trail. If flagged, x402-secure blocks the payment and keeps the spending limit intact.


MCP marketplace integration


t54 also tested the same integration against the Coinbase x402 Bazaar, a marketplace of paid AI tool servers (Model Context Protocol, or MCP, servers). The agentic system connects through Amazon Bedrock AgentCore Gateway, a capability of Amazon Bedrock AgentCore, discovers paid tools, and calls them. When a tool returns an x402 payment requirement, ProcessPayment signs the transaction. One Amazon Bedrock AgentCore payments setup covers both direct API endpoints and marketplace-listed tools without additional configuration.


The trust signals


Table 1 lists the x402-secure endpoints available for scoring before payment.

































Endpoint What it checks
POST /x402/tools/get_overall_score Comprehensive security score with risk indicators
POST /x402/tools/get_onchain_trust Blockchain address risk detection
POST /x402/tools/get_webpage_trust AI-powered phishing and unauthorized activity detection
POST /x402/tools/get_social_trust Social media reputation scoring
POST /x402/tools/get_api_health Server reliability and compliance
POST /x402/tools/evaluate_agent_payment Pre-transaction payment risk evaluation (Base)

Table 1: x402-secure trust-scoring endpoints


Observability and audit


Every ProcessPayment call emits structured logs (session, instrument, amount, status) to Amazon CloudWatch. AWS CloudTrail captures the full API history for compliance review. Trust decisions and payment outcomes are correlated in Amazon CloudWatch Application Signals on a per-session basis. For regulated workloads, this gives compliance teams a continuous audit trail of every dollar an agent spends and the trust signal it spent against.


Get started


If you are building agents that hit paid APIs:


The walks through your first agentic payment end-to-end. Add trust-scored payments to an existing AgentCore agent with the open-source SDK:



BASH
pip install x402-secure


If you are an API provider offering paid endpoints:


Replace your x402 facilitator URL with the t54 proxy to add trust protection. Same x402 API, one configuration change:



TEXT
Before: https://x402.org/facilitator
After: https://x402-proxy.t54.ai


Learn more



  • Read the launch blog, for a step-by-step walkthrough of your first agentic payment.

  • Explore the for the full API reference and configuration guide.

  • Visit the , which is open source under the MIT license.

  • View the .




About the authors





Chris Wajule



Chris Wajule


is a Principal Product Manager – Technical at AWS with 12+ years of experience in product and business management. He is passionate about building and delivering technology products that create meaningful impact in customers’ lives.





Chandler Fang



Chandler Fang


is a founding team member at t54 and led the integration of x402-secure with Amazon Bedrock AgentCore payments described in this post. His work focuses on the protocol and risk-scoring infrastructure that allows agents to transact autonomously under verifiable guardrails.



Vollständiger Original-Bericht
Ausführliche Details, Code-Beispiele & Hersteller-Stellungnahme auf aws.amazon.com.
↗ Original-Artikel auf aws.amazon.com lesen
Wie bewertest du diesen Beitrag?
1 Klick Feedback
85 Fachleser & IT-Security Experten haben diesen Report heute geteilt
Teilen mit Netzwerk & Team:

Community-Analysen & Experten-Meinungen 0

Verfasse deine eigene Analyse, teile Workarounds oder diskutiere diesen Vorfall im Blog.
Noch keine Community-Analyse verfasst. Markiere einen Textabschnitt oder klicke oben auf Eigene Analyse verfassen“!
Community Pulse: Relevanz-Einschätzung
1 Klick Experten-Votum
🔴 Akute Relevanz 44%
🟡 In Evaluierung 22%
🟢 Keine Auswirkung 12%
Spannende Innovation 22%
Verwandte Story-Cluster & Quellen (Vektor-KI)
Port 8095 Engine
4 Quellen
DSA-6477-1 linux - security update
1 Quelle
KI-Bilderkennung: Wie funktioniert Image Recognition?
1 Quelle
Angreifer kapern Zimbra-Server über SNMP-Lücke