Many enterprise workflows depend on desktop applications—ERP systems, internal tools, and proprietary software—that expose no APIs. AgentBay Computer Use gives AI agents a cloud desktop environment where they can see the screen, click, type, and run full automation flows, just as a human operator would. Agents connect via SDK or MCP (Model Context Protocol) and run on AgentBay's multi-modal cloud environment.
Use cases
-
Computer application testing
Functional testing: Simulate user interactions to verify that desktop application features work correctly.
Compatibility testing: Run the same desktop application against multiple operating system images to confirm cross-environment behavior.
Performance testing: Measure response times and resource usage under simulated user load.
-
Office automation
Document processing: Create, edit, and convert documents without human intervention.
Email management: Receive, categorize, draft replies, and send emails automatically.
Office communication: Process instant messages, summarize threads, and generate responses.
Report generation: Compile data and produce business and analysis reports on a schedule.
Workflow approval: Route and approve office workflows that would otherwise require manual sign-off.
-
System administration
System configuration: Apply and verify system settings across environments automatically.
Software installation: Deploy and configure software applications in bulk.
Maintenance tasks: Run cleanup and maintenance routines without scheduling human downtime.
-
Research and education
Experiment control: Automate the control of scientific equipment and experimental procedures.
Data processing: Collect, clean, and analyze experimental data automatically.
Teaching assistance: Organize and distribute learning materials to students at scale.
Core capabilities
-
Efficient and flexible
Multi-platform support: Supports multiple Windows and Linux image versions, plus custom images, to match the runtime requirements of any desktop application.
Elastic scheduling: Creates, assigns, and releases compute resources on demand so agents can scale up or down without manual intervention.
Environment isolation: Each task runs in its own independent image, keeping runtimes stable and preventing tasks from interfering with each other.
-
Enterprise-grade security
Network isolation: Uses Virtual Private Cloud (VPC) to isolate agent traffic from the public internet and from other tenants.
Data encryption: Encrypts data at rest and in transit to meet enterprise compliance requirements.
Access control: Provides fine-grained computer access control and supports network and application allowlists to restrict which resources an agent can reach.
-
Developer-friendly
Multiple access methods: Connect via SDK, MCP, or the ASP streaming protocol. Python, TypeScript, and Golang are all supported.
Visual debugging: The ASP streaming protocol streams a live view of what the agent sees. Session recordings let you replay and diagnose automation flows after the fact.
Serverless architecture: Built on a cloud-native serverless model with automatic scaling and pay-as-you-go billing, so you pay only for what your agents use.