Microsoft has introduced CodeAct, an enhancement to its .NET Agent Framework, designed to significantly speed up AI agents. Instead of making multiple sequential calls to an AI model for different tools, CodeAct allows the model to generate a single Python program that executes all necessary steps at once. This approach reduces latency by approximately 50% and cuts token consumption by over 60% on representative workloads. The system utilizes Hyperlight, a micro-virtual machine technology, to securely sandbox the AI-generated code, ensuring it only interacts with explicitly authorized tools. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Accelerates AI agent execution by consolidating tool calls into a single program, reducing latency and costs.
RANK_REASON This is a product update to an existing framework, offering performance improvements and new features for AI agents.