Amazon WorkSpaces for AI agents FAQs

Amazon WorkSpaces gives your AI agents secure, managed desktop environments to operate the enterprise applications your business depends on. Agents connect through a standard Model Context Protocol (MCP) endpoint and interact with desktop application interfaces, seeing the screen, clicking, and typing the way a human would, without requiring API access or application modernization.

Critical business processes on desktop applications (ERP systems, CRMs, mainframes, and proprietary tools) where years of customization and strict compliance requirements make them costly to modernize. WorkSpaces gives your AI agents a managed cloud desktop environment where they can see the screen and take action within these applications similar to workers do today, without requiring custom integrations.

Agents inherit the same identity controls, network isolation, and compliance boundaries as human users. Organizations can automate workflows such as claims processing, patient record updates, trade settlement, and back-office operations while maintaining existing governance policies.

WorkSpaces for AI agents is available in: US East (N. Virginia, Ohio), US West (Oregon), Canada (Central), EU (Frankfurt, Ireland, London, Paris), and Asia Pacific (Mumbai, Singapore, Sydney, Tokyo, Seoul).

For the latest availability, see the allowed domains documentation.

Agents interact with applications using a combination of vision-driven action and MCP tools:

• MCP tool forwarding exposes multiple tool paths so your agent framework can route each subtask to the most efficient option available.
• Domain join lets you connect agent sessions to your Active Directory, enforcing the same identity and access policies your organization already manages.
• Over-the-shoulder monitoring gives operators real-time visibility into agent sessions so you can observe, audit, or intervene as needed.

For more details, see the MCP server documentation.

WorkSpaces for AI agents is a bring-your-own-agent model. It connects through a standard MCP endpoint using the AI framework and model you already run. No proprietary SDKs, no rearchitecting your agent infrastructure.

Your model and inference costs are separate and not included in WorkSpaces for AI agents pricing. WorkSpaces provides the managed desktop environment and MCP connectivity; you provide the agent.

For compatible frameworks and configuration details, see the MCP server documentation.

You can connect an agent to a desktop application in under 15 minutes. Enable access for your agents on a new or existing stack, generate a streaming URL, pass it to your MCP client, and the agent connects to the session.

You can configure this through the Amazon WorkSpaces Applications console, AWS CLI, or API. For domain-joined fleets, use a signed SAML assertion instead of a streaming URL.

For a step-by-step walkthrough, see the getting started guide or the sample-code-for-workspaces-agent-access repository on GitHub.

For information on Microsoft 365 unattended licensing for agent-based automation, see the AWS and Microsoft FAQ.

WorkSpaces for AI agents pricing has two components:

• WorkSpaces Applications instance pricing — The streaming instance hosting your agent's desktop environment is billed hourly based on instance type.
• Agent access MCP session fee — Billed per second of active session time. A session starts when your agent connects and ends when it disconnects or goes idle. There are no per-call fees; make as many tool calls as you need during a session without incurring additional charges.

There are no upfront commitments or long-term contracts. If you run Always-On fleets, agent session fees apply only when the agent is actively connected, not for the full duration of your fleet instance.

Every account includes 10 free hours of session time per region each month. After you've used your free hours, standard per-hour rates apply. Free tier hours reset monthly and don't roll over.

For full pricing details, see the WorkSpaces Applications pricing page.