What is the Model Context Protocol?
An open standard by Anthropic defining how AI agents connect to external tools and data sources. MCP provides a universal interface for discovering and invoking capabilities — tools (actions), resources (data), and prompts (interaction patterns) — across any compatible agent framework.
WHY IT MATTERS
Before MCP, every framework had its own tool integration — custom APIs, incompatible formats. MCP standardises this with a universal protocol that any agent can speak.
Three primitives define the protocol: tools (executable actions an agent can perform), resources (read-only data providing context), and prompts (reusable interaction templates). Together they cover the full surface area of agent-to-system interaction.
MCP is rapidly becoming the standard for agent tool access. Claude, Cursor, Windsurf, VS Code Copilot, and most major agent frameworks now support it natively — meaning any MCP-compatible tool is automatically available to any MCP-compatible agent.
HOW POLICYLAYER USES THIS
PolicyLayer is a hosted MCP gateway that sits transparently between any MCP client and server. Rather than modifying agent code or server implementations, you point the client at PolicyLayer instead of the server. PolicyLayer evaluates every tool call against YAML-defined policies — enforcing allow/deny decisions, argument constraints, and rate limits — then forwards approved calls to the upstream server. No code changes needed, fail-closed by default.
IN THE CATALOGUE
PolicyLayer continuously scans the MCP ecosystem and classifies every tool it finds by risk category.