MarketNow Launches MCP Trust Layer to Secure AI Agent Commerce

MarketNow has launched a "trust layer" for the Model Context Protocol (MCP) ecosystem, providing a secure registry for AI agent skills. The platform aims to bridge the gap between tool discovery and security verification as autonomous agents gain more autonomy.

To address security risks, MarketNow introduced "Sentinel L2.5," a scanning system that performs real-time security audits of skills. Sentinel employs 18 Semgrep rules, OSV, and Docker/gVisor sandboxing to detect and block malicious activity. Alongside these security measures, the platform provides a registry of AI agent skills, including a subset of 43 free options.

The Trust Gap in MCP

The Model Context Protocol is an open standard designed to allow AI agents to connect seamlessly to external data sources and tools. While existing registries like Smithery and Glama have largely solved the problem of discovery—helping users find the tools they need—the industry has struggled with "trust." Verifying that a third-party tool is not malicious remains a significant hurdle for developers and enterprises deploying agents in production environments.

Securing Agent Commerce

As agents move toward executing financial transactions, MarketNow has integrated a payment infrastructure supporting credit cards via Stripe and USDC on the Base L2 network through the x402 protocol. To prevent unauthorized spending, the system implements "AP2-compatible mandates," which ensure that human-in-the-loop approval is required for agent purchases by default.

This layer of interception is critical as the risk of prompt injection increases. Without such safeguards, malicious prompts could potentially lead to data exfiltration—such as the unauthorized reading of .env files—or unplanned financial expenditures. A standardized trust layer that can block dangerous commands in real-time is seen as a prerequisite for the broader enterprise adoption of autonomous agents.

Looking Ahead

While the functional claims regarding Sentinel and the x402 protocol are central to the launch, the platform's exact registry size has varied across sources. Additionally, early community feedback on Hacker News indicated that the project's linked public GitHub repository was returning 404 errors at the time of posting. Observers will be watching to see if MarketNow can establish itself as the industry standard for agent verification as the MCP ecosystem continues to scale.