As reported by Silicon Angle, Manifold Security Inc. has expanded its Manifest supply chain intelligence tool to include Model Context Protocol (MCP) servers, aiming to address a growing security gap in AI agent integrations.
The expansion adds scored entries for over 7,700 MCP servers to Manifold's existing index of agent skills and plugins. Each server receives a Manifest Score, derived from a Lineage Score evaluating publisher provenance and a Safety Score flagging behavioral risks. MCP, developed by Anthropic, is crucial for connecting AI agents to external resources, but the security review process for individual servers has lagged behind its widespread adoption. Unlike agent skills often backed by public repositories, many MCP servers expose only an HTTP endpoint, making their security difficult to verify. A compromised MCP server can control an agent's reasoning, execution, and data flow, making provenance and behavioral signals critical for enterprises.
The Lineage Score assesses authorship history, community presence, and repository details, while the Safety Score inspects the server's declared interface for contradictions or manipulative content, such as prompt injection. Manifold's tool now indexes over 206,000 assets, with plans for further coverage.
Source: Silicon Angle
