Most automation runs against fixed or manually configured environments. You build against one browser, one OS, and one screen size. When that setup drifts, your script breaks. The /v1/machines API lets you provision cloud machines on demand so your computer use agent always runs against a fresh, real desktop, browser, or terminal.
How it works
You send a POST to /v1/machines with machine_id, os, and size. The server creates a cloud VM and returns a state object. You can start, stop, and snapshot the machine. Your agent then connects to the running desktop and drives it with computer use actions, starting from the vision endpoint /v1/predict or the stateful session endpoint /v1/sessions/{id}/predict. The machine stays live until you stop or cancel it, giving you a persistent environment for complex workflows.
curl https://coasty.ai/v1/machines \
-H "X-API-Key: $COASTY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"machine_id": "my-test-box",
"os": "ubuntu-latest",
"size": "large"
}'Machine lifecycle
- POST /v1/machines creates a cloud VM with a unique machine_id
- GET /v1/machines lists your provisioned machines and their state
- You can start and stop machines to manage costs
- Use POST /v1/machines/{id}/stop to shut down a machine
- Use POST /v1/machines/{id}/snapshot to capture a state for reuse
- The agent drives the running desktop with computer use actions
- Billing is based on the agent steps you run on the machine, not the machine itself
POST /v1/machines creates a cloud VM. Your computer use agent runs against the running desktop.
Where this beats brittle automation
API-only tools depend on stable endpoints and documented payloads. If a site changes a selector, breaks a GraphQL route, or adds a modal, your script fails. With a computer use agent on a real machine, the agent sees the screen, understands the context, and adjusts clicks and inputs like a human. It can handle dynamic content, pop-ups, and layout shifts without brittle selectors. You also get a live browser or terminal that behaves exactly as a user would, not a mocked environment.
Start provisioning cloud machines on demand and let the computer use agent handle the rest. Spin up fresh environments, run complex workflows, and iterate faster. Get your API key at https://coasty.ai/developers and build agents that work on any UI.
Want to see this in action?
View Case Studies