v3 vs v4: Choosing a Computer Use Model on the API
You want an agent that can click, type, and browse like a human. The Coasty Computer Use API gives you two model flavors via the cua_version parameter. v3 is a step-by-step driver you control. v4 is autonomous with a built-in verifier. This guide shows how they differ, when to pick each, and how to call the Task Runs endpoint correctly.
How it works
Use POST /v1/runs to start a task run. The cua_version field selects the model. v3 (default) runs step-by-step. v4 adds an autonomous verifier that checks success or failure. You provide machine_id (the cloud VM), task (what the agent should do), and optional instructions and system_prompt. The server bills $0.05 per agent step. You can stream events with GET /v1/runs/{id}/events to watch progress. States include queued, running, awaiting_human, succeeded, failed, cancelled, and timed_out.
#!/bin/bash
# Start a v3 task run with a cloud machine
COASTY_API_KEY=$(cat ~/.coasty_key)
machine_id="m-12345"
curl -s https://coasty.ai/v1/runs \
-H "X-API-Key: $COASTY_API_KEY" \
-H "Content-Type: application/json" \
-d "{
\"machine_id\": \"$machine_id\",
\"task\": \"Open Chrome, navigate to coasty.ai, and click the Developers link\",
\"cua_version\": \"v3\",
\"instructions\": \"Be concise and use the visible UI elements.\",
\"max_steps\": 50,
\"deadline_seconds\": 600
}" | jq .v3 model overview
- ●The default cua_version value
- ●Step-by-step control flow
- ●You decide when to pause or cancel
- ●Perfect for workflows needing human-approval steps
- ●Billed $0.05 per agent step
v4 model overview
- ●Set cua_version to v4 for autonomous mode
- ●Includes a built-in pass/fail verifier
- ●Agent drives itself to completion or failure
- ●Ideal for fully automated tasks
- ●Still billed $0.05 per agent step
Set cua_version to v4 for autonomous mode with an internal verifier, or keep the default v3 for step-by-step control.
Where this beats brittle automation
Traditional automation relies on brittle selectors and fixed API endpoints that break when UI changes. A computer use agent sees the screen, understands context, and uses visible elements. It can handle dynamic layouts, pop-ups, and mixed interfaces without brittle selectors. You get higher reliability and less maintenance when the UI changes.
Start with v3 for controlled, step-by-step tasks, then try v4 for fully automated runs. Use GET /v1/runs to list runs and GET /v1/runs/{id}/events to stream progress. Get a key at https://coasty.ai/developers and start building your computer use agents today.