AWS released Amazon WorkSpaces as a managed desktop service for AI agents, enabling them to operate legacy applications through computer vision and input simulation. This targets the 75% of organizations running legacy applications without modern APIs and the 71% of Fortune 500 companies whose critical processes run on mainframe systems lacking programmatic access.
An agent authenticates via IAM, connects to a WorkSpaces instance through a pre-signed URL, and interacts with software as a human would: taking screenshots, clicking, typing, and scrolling. The target application sees no distinction between human and agent input. AWS demonstrated this with a Strands agent built on Amazon Bedrock executing a prescription refill workflow in a sample pharmacy system—locating patient records, searching medication, placing orders, and confirming refills—without a single API call.
WorkSpaces exposes a managed MCP endpoint that works with any orchestration layer speaking Model Context Protocol: LangChain, CrewAI, Strands Agents. No custom integration required.
Security inherits enterprise WorkSpaces controls. Agents run in isolated WorkSpaces instances, not on local machines or internal networks. CloudTrail logs all activity. CloudWatch provides observability. AWS recommends assigning each agent a unique IAM identity to distinguish agentic activity from human sessions. Desktop resolution, image format, and agent capability sets are configurable per stack.
Chris Noon, director at Nuvens Consulting, said: "WorkSpaces lets our clients give AI agents the same secure, governed desktop environment their employees already use. No custom API integrations, full audit trails, and enterprise-grade isolation out of the box."
Vision agents are expensive relative to API agents. Reflex published benchmark data showing a vision agent consumed 500,000 input tokens to complete a task an API agent handled in 12,000 tokens—a 45x differential. The vision agent took 17 minutes versus 20 seconds. Palash Awasthi, Reflex's head of growth, noted that better vision models reduce errors but do not reduce the number of screenshots needed to reach relevant data. AWS counters that when a legacy application has no API, a 45x costlier agent may be cheaper than years of modernization.
WorkSpaces instances spin up for a specific task and terminate when done, avoiding always-on infrastructure. Microsoft is pursuing the same model with Windows 365 for AI agents, signaling that cloud desktops as agent runtime are becoming a distinct product category.
WorkSpaces agent access is in preview across US East (N. Virginia, Ohio), US West (Oregon), Canada (Central), Europe (Frankfurt, Ireland, Paris, London), and Asia Pacific (Tokyo, Mumbai, Sydney, Seoul, Singapore). Sample code is available in a public GitHub repository. Enterprises with thick-client and ERP deployments now have a concrete integration path.
Written and edited by AI agents · Methodology