Amazon WorkSpaces for AI agents (Preview)
Secure, managed desktop environments for AI agents to operate enterprise applications
Why WorkSpaces?
Amazon WorkSpaces lets AI agents take action on the desktop applications your business depends on. Connect your agents through a standard MCP endpoint without requiring API access or application modernization.
Benefits
Point your agent framework at a managed WorkSpaces endpoint and start automating. Your agents interact directly with desktop application interfaces, so you skip the months of API development and application rework that custom integrations require.
WorkSpaces combines vision-driven desktop interaction with MCP tool forwarding to route each subtask to the most efficient interface. When purpose-built MCP tools are available, agents use them for speed and cost efficiency, falling back to vision-driven action only when necessary. This reduces latency and improves reliability across complex workflows.
Connect through a standard MCP endpoint using the AI framework and model you already run. No proprietary SDKs, no vendor lock-in, no rearchitecting your agent infrastructure.
Agents assume known identities within your Active Directory through domain joined fleet support, so the same fine-grained access control policies you enforce for human users apply to every agent session. Every action is logged and auditable through your existing compliance workflows.
How it works
Connect your agent
Your agent connects to desktop applications through a managed WorkSpaces Applications environment and a standard MCP endpoint. Authenticate through IAM, point your agent framework at the endpoint, and the agent accesses a Windows desktop configured with the applications you specify.
Agents take action on the desktop
Agents interact with desktop applications using a combination of vision-driven action and MCP tools. WorkSpaces routes subtasks to the most efficient path available, using MCP tools when they exist and falling back to screen-level interaction when they don't.
Stay in control
Maintain oversight of agent sessions with real-time user control. Observe agent actions as they happen with view-only access, or intervene mid-operation to pause and stop a session when needed. Configure the level of visibility appropriate for each workflow, whether during development, testing, or production. AWS CloudTrail logs every tool call, Amazon CloudWatch tracks session metrics, and screenshots are stored in Amazon S3 for debugging and audit.
Use cases
How teams put AI agents to work on desktop applications
Back-office operations
Enable your agents to handle invoice entry, record reconciliation, and cross-application data transfers to free your operations teams from repetitive manual work.
Claims and compliance
Have your agents cross-reference claims data across multiple desktop applications, validate entries against policy rules, and compile audit-ready submission packages, replacing the manual comparisons, copy-paste, and documentation that slow down compliance teams today.
Workforce management
Connect your AI agents to WorkSpaces to handle transfers between HR platforms, provisioning tools, and benefits systems, keeping employee records consistent without manual re-entry.