Provision Cloud Machines on Demand with the /v1/machines Computer Use API
Most automation stacks rely on a single long‑lived desktop or a pool you manage yourself. Adding new environments means updating scripts, waiting for provisioning, and juggling SSH keys and connections. The /v1/machines endpoint changes that by letting you provision fresh cloud desktops on demand. You can spin up isolated environments, run a computer use agent, then tear them down when the task finishes. This removes operational overhead and lets you focus on the agent logic itself.
How it works
You send a POST request to /v1/machines with a JSON payload that describes the machine. The response returns an ID you use later to start, stop, or snapshot that machine. The API does not expose low‑level VM details like OS version or region. It abstracts those choices and ensures the environment is ready to accept a computer use agent. This design keeps the interface simple and lets you iterate quickly on agent behavior without infrastructure changes.
curl -X POST https://coasty.ai/v1/machines \
-H "Authorization: Bearer $COASTY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"machine_id": "my-prod-tester",
"task": "run a browser automation task",
"cua_version": "v3"
}'Key request fields
- ●machine_id: a human‑readable identifier for the machine (required).
- ●task: a short description of what should happen on this machine (required).
- ●cua_version: the computer use agent version to run on the machine (default 'v3').
Machine lifecycle
- ●After provisioning you receive an id that you use with GET /v1/machines to see the current state.
- ●Start the machine with POST /v1/machines/{id}/start.
- ●Stop the machine with POST /v1/machines/{id}/stop.
- ●Take a snapshot with POST /v1/machines/{id}/snapshot to save the current desktop state.
POST /v1/machines creates a cloud machine, then you control it with POST /v1/machines/{id}/start, /stop, and /snapshot.
Where this beats brittle automation
Traditional automation often relies on CSS selectors, XPath, or hardcoded indices to find elements on a page. Changes in layout or styling break these selectors and force you to update code frequently. A computer use agent can see the screen, interpret instructions, and act like a human. It can click or type on any element it recognizes, regardless of implementation details. This makes your automation more resilient to UI changes and reduces maintenance overhead.
Use the /v1/machines endpoint to spin up dedicated desktop environments for each run. Run a computer use agent on them, and tear them down when you are done. This pattern scales well for automated testing, QA pipelines, and any scenario where you need an isolated, real‑desktop environment. Get a key at https://coasty.ai/developers to start provisioning machines on demand.