Cursor AI Updates Cloud Agents with Up to 80% Token Efficiency Gains
Cursor has updated its cloud agents to improve token efficiency by 20-30% overall. Additionally, tasks involving computer use and Model Context Protocol (MCP) integration see an efficiency improvement of up to 80%. This update allows developers to delegate more complex, long-running tasks to AI agents while staying within budget, reducing the high API costs typically associated with agentic workflows. The efficiency gains are achieved through optimized handling of MCPs, skills, and computer use interactions. This allows the agents to execute actions and retrieve external context with significantly fewer tokens.
## BACKGROUND
Cursor is a popular AI-powered code editor built on top of Visual Studio Code. The Model Context Protocol (MCP) is an open-source standard designed to connect AI models to external data sources and tools, while "computer use" refers to an AI's ability to interact directly with operating systems and applications.