Microsoft Open-Sources Agent Control Specification and SDK with Plugins for LangChain, AutoGen, OpenAI, Anthropic and Other Agent Frameworks

Microsoft released a new open-source standard called Agent Control Specification, or ACS, that lets developer, compliance and security teams define their own policies for agents to follow in portable policy files. ACS policies can define what the agent may do, what it must not do, when a human should approve an action, and what evidence should be logged for later review.

The policies are checked at several interception points when the agent is performing a task.

Checks occur before an agent receives input, before it calls a tool, after a tool returns a result, and before the final response is sent to the user. An ACS policy may allow an action, block it, redact sensitive information, or ask a person to approve it.

ACS supports insertion of classifiers for inputs and outputs to categorize information, predict outcomes, or determine how an agent should respond.

It also supports adding LLMs with prompts to act as a judge for policies and logic for checking tool calls, tool selection, input accuracy, output usage, and responses. ACS policy files can be bundled with agents so a security policy follows the agent across different frameworks and environments. AI and MCP tools.