Build a Self-Running QA Testing Bot With the Computer Use API
QA teams often rely on brittle selectors and fragile APIs to test workflows. When the UI changes or an element shifts, tests break. You can fix this by building a self-running QA testing bot that sees the screen and acts like a human. With Coasty's computer use API, you provision a cloud VM, send a task, and get a live desktop driven by an agent that captures screenshots, predicts clicks and types, and asserts outcomes. No hardcoded CSS selectors, no fragile API mocks, just a real browser or desktop experience that succeeds or fails in the same way a human does.
How a QA bot works with Coasty
Your QA bot runs as a Task Run on a provisioned machine. You POST to /v1/runs with a machine_id (a cloud VM you provision via /v1/machines), a task describing the QA flow, a cua_version (v3 for guided, v4 for autonomous with a pass/fail verifier), and optional instructions. The server launches an agent that loops: capture a screenshot, run prediction, perform actions, and repeat until status is done. You can stream events with GET /v1/runs/{id}/events to show progress. The agent bills at $0.05 per step. After the run completes, you check status: succeeded or failed.
curl https://coasty.ai/v1/runs \ \
-H "X-API-Key: $COASTY_API_KEY" \ \
-H "Content-Type: application/json" \ \
-d '{
"machine_id": "machine_123",
"task": "Navigate to https://example.com, verify the title contains 'Example Domain', and click the first link.",
"cua_version": "v4",
"on_awaiting_human": "pause",
"max_steps": 50
}'Provision a cloud machine for your QA environment
- ●POST /v1/machines creates a cloud VM with a unique machine_id.
- ●You start and stop the machine as needed to run isolated QA sessions.
- ●The agent drives real desktops, browsers, and terminals, not just API calls.
Provision a machine with POST /v1/machines, then drive QA runs on that machine_id.
Use workflows to orchestrate complex QA checks
For multi-step QA flows, define a workflow as a versioned JSON DSL and POST /v1/workflows. Each step can be a task, assert, if, loop, parallel, human_approval, retry, succeed, or fail. Hard guards like budget_cents, max_iterations, and deadline_seconds keep tests safe. Task steps incur $0.05 each. Workflow steps let you chain assertions, retry failing checks, and parallelize independent tests across machines.
Where this beats brittle automation
Traditional QA tools rely on CSS selectors or XPath, which break when the UI changes. A computer use agent sees the actual screen, recognizes buttons and links by visual context, and performs actions like a human. It can handle dynamic content, pop-ups, and layouts that resist selector-based automation. With Coasty's autonomy options (v4 with a pass/fail verifier), you get one verification call instead of manually asserting every state. This makes testing more resilient and easier to maintain.
You can now spin up a self-running QA testing bot that drives real desktops and browsers with a computer use API. Start by provisioning a machine with POST /v1/machines, then run QA tasks with POST /v1/runs or orchestrate complex flows with a workflow. Get your API key at https://coasty.ai/developers and build your first autonomous QA bot today.