What is an Agent Framework?
An agent framework is a software library that provides abstractions for building AI agents, handling concerns like tool management, state persistence, conversation flow, multi-agent coordination, and LLM integration.
WHY IT MATTERS
Building an AI agent from scratch means solving tool calling, state management, error handling, memory, and orchestration yourself. Agent frameworks handle these concerns so developers can focus on business logic.
The landscape has matured rapidly. LangGraph offers stateful graph-based agents. CrewAI specialises in multi-agent teams. OpenAI's Agents SDK provides a minimal, production-focused approach. PydanticAI emphasises type safety. Each makes different tradeoffs between flexibility and simplicity.
Critically, most major frameworks now support MCP natively — meaning they can connect to MCP servers (and MCP proxies like PolicyLayer) out of the box. This decouples agent development from policy enforcement: build your agent in whatever framework you prefer, then add policy governance by pointing it at PolicyLayer.
HOW POLICYLAYER USES THIS
PolicyLayer works with any agent framework that supports MCP — which includes Claude Desktop, Cursor, Windsurf, VS Code Copilot, LangChain, OpenAI Agents SDK, PydanticAI, and more. Because PolicyLayer is a transparent MCP proxy, no framework-specific plugin or SDK is needed. Point the framework's MCP client at PolicyLayer, define your YAML policies, and tool calls are governed automatically.