You need an agent that watches your screen and clicks, types, and scrolls just like a human. The Coasty Computer Use API lets you drive real desktops, browsers, and terminals. Two model versions are available. v3 is a controlled, step-by-step loop you manage. v4 runs autonomously with a built-in pass/fail verifier. This post explains the real differences, request shapes, and when to pick each version.
How it works
Task runs start with a POST to /v1/runs. You provide a machine_id, a task description, and the cua_version you want to use. The server provisions a cloud VM and launches the agent. v3 (default) returns actions one step at a time. v4 (autonomous) runs until the verifier decides success or failure. The API bills $0.05 per agent step in both cases. You can stream events with GET /v1/runs/{id}/events to see progress in real time.
#!/usr/bin/env bash
export COASTY_API_KEY
# Start an autonomous run with v4
# cua_version: v3 (controlled) or v4 (autonomous with verifier)
curl -X POST https://coasty.ai/v1/runs \
-H "X-API-Key: $COASTY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"machine_id": "machine-123",
"task": "Open the Coasty website and verify the navigation bar is visible.",
"cua_version": "v4",
"max_steps": 50,
"deadline_seconds": 300,
"on_awaiting_human": "pause"
}'
Key differences between v3 and v4
- cua_version: set to "v3" for controlled, step-by-step execution, or "v4" for autonomous runs with a pass/fail verifier.
- Control flow: v3 requires you to capture the screen, predict, and act in a loop. v4 lets the server drive the agent to completion.
- Verifier: v4 includes an automated verifier that evaluates whether the task succeeded or failed.
- Cost: both versions are billed $0.05 per agent step when using task steps.
- States: both can reach queued, running, awaiting_human, succeeded, failed, cancelled, or timed_out.
Pick v3 when you need detailed control over each step. Pick v4 when you want an autonomous run with a built-in pass/fail verifier.
Where this beats brittle automation
Traditional automation relies on brittle selectors, hardcoded IDs, and fragile API endpoints. If a UI changes, your scripts break. A computer use agent sees the screen like a human. It handles layout shifts, dynamic content, and unexpected error states. Coasty agents drive real desktops and browsers using computer vision and natural language instructions. This means fewer maintenance cycles and more robust workflows, especially for complex or long-lived tasks.
Choose the right computer use model for your project. v3 gives you step-by-step control. v4 provides autonomous execution with a verifier. Start building reliable agents that see and act like humans. Get your API key at https://coasty.ai/developers.
Want to see this in action?
View Case Studies