|
Enterprises face a big problem when deploying AI brokers: the desktop and legacy purposes that energy most enterprise workflows are merely inaccessible to fashionable AI methods. In line with a 2024 Gartner report, 75% of organizations run legacy purposes that lack fashionable APIs, and 71% of Fortune 500 firms function essential processes on mainframe methods with out enough programmatic entry. For a lot of organizations, this has meant selecting between delaying AI adoption or enterprise costly and dangerous modernization initiatives.
Immediately, we’re saying that Amazon WorkSpaces now allows AI brokers to securely function desktop purposes with out requiring utility modernization. The identical managed digital desktops that thousands and thousands of workers use and belief can now additionally serve AI brokers, turning WorkSpaces into infrastructure for scaling enterprise productiveness, not simply delivering it. As a result of brokers function inside your current WorkSpaces setting, there are not any APIs to construct, no utility migrations to plan, and no new infrastructure to handle.
A few of our prospects had an early alternative to provide their brokers a WorkSpace. Chris Midday, Director, Nuvens Consulting shared with us, “WorkSpaces lets our shoppers give AI brokers the identical safe, ruled desktop setting their workers already use — no customized API integrations, full audit trails, and enterprise-grade isolation out of the field. For regulated industries, that’s not a nice-to-have — it’s the baseline.”
Safe cloud desktop entry for AI brokers
With WorkSpaces, AI brokers can securely entry and function desktop purposes operating inside managed WorkSpaces environments to finish complicated enterprise workflows. Brokers authenticate via AWS Identification and Entry Administration (IAM) and join through Workspaces with full audit trails accessible via AWS CloudTrail and Amazon CloudWatch. As a result of brokers function inside safe WorkSpaces environments reasonably than on native machines, your current safety controls and compliance insurance policies stay absolutely intact.
Amazon Workspaces helps the industry-standard Mannequin Context Protocol (MCP), which suggests WorkSpaces works with any agent framework, corresponding to LangChain, CrewAI and Strands Brokers.
Let’s attempt it out
To arrange a WorkSpaces setting for AI brokers, I began within the AWS Administration Console by creating a brand new WorkSpaces Purposes stack—the setting definition that controls how brokers join and what they’re allowed to do.
From the Amazon WorkSpaces console, I selected Create stack and configured the fundamentals: identify, fleet affiliation, and VPC endpoints. In Step 3 of the stack creation workflow, I observed the brand new AI brokers part with two choices. The primary, No AI agent entry, is the default configuration for normal WorkSpaces designed for folks. The second, Add AI Brokers, permits AI brokers to securely entry and function purposes utilizing their very own identification and permissions. I chosen Add AI Brokers to allow agent connections on this stack.
Subsequent, I’ll allow storage earlier than configuring the agent entry settings to outline how brokers work together with the desktop.
Beneath Agent options, I enabled three capabilities. Laptop enter permits the agent to click on, kind, and scroll throughout the desktop. Laptop imaginative and prescient permits the agent to seize screenshots of the desktop, which is the way it “sees” the applying. Lastly, screenshot storage configures the place session screenshots are saved for audit and debugging.
Beneath Desktop display screen format, I set the display screen decision to 1280×720 and picture format to PNG. The decision determines the constancy of what the agent sees throughout a session—a posh utility with dense UI parts would possibly profit from increased decision, whereas a terminal-style interface works properly at 720p.
With my stack configured, WorkSpaces exposes a managed MCP endpoint. I pointed my agent framework to this endpoint, supplied IAM credentials for authentication, and my agent started interacting with the desktop purposes put in on the fleet’s picture.
To see this in motion, right here’s an agent constructed with the Strands Agent SDK and Amazon Bedrock dealing with a prescription refill, trying up the affected person file, trying to find the treatment, inserting the order, and confirming a profitable refill, all inside a pattern pharmacy system with no API.
The appliance doesn’t know an agent is driving it. Nothing concerning the software program was modified, rebuilt, or built-in. The agent labored with it precisely because it exists at this time.
Now accessible
This function is accessible at this time in public preview at no extra value in US East (N. Virginia, Ohio), US West (Oregon), Canada (Central), Europe (Frankfurt, Eire, Paris, London), and Asia (Tokyo, Mumbai, Sydney, Seoul, Singapore) Areas.
Get began constructing at this time utilizing our GitHub repo, or go to the WorkSpaces web page for extra particulars.
