Decawork provides policy enforcement and approval systems for AI agents. Companies set policies in plain English, and Decawork trains a company-specific model to keep agent actions aligned with those policies. The platform routes all API and MCP calls, enforces real-time approval based on context, and maintains an agent registry.