Coasty Computer Use API Pricing: Every Endpoint, Every Cent
You do not want to ship a computer use agent and discover you are paying $0.20 per task. You also do not want to guess what the /v1/predict endpoint costs or how billing works for workflows. This post lists every Coasty endpoint with its exact price and request/response fields. You will see why a computer use agent that sees the screen and acts like a human is cheaper and more robust than brittle selectors and API-only tools.
Base setup and credentials
- ●Base URL: https://coasty.ai/v1
- ●Auth header X-API-Key: <key> or Authorization: Bearer <key>
- ●Read the key from COASTY_API_KEY env var, never hardcode it
- ●Docs: https://coasty.ai/docs , keys: https://coasty.ai/developers/keys
Vision endpoints
- ●POST /v1/predict , take a screenshot, instruction, cua_version. Returns actions and a status. Loop capture, predict, act until status is done. Price: $0.05 per call.
- ●POST /v1/sessions , start a stateful trajectory session. Returns an id. Price: $0.10 per session.
- ●POST /v1/sessions/{id}/predict , in a session, send a screenshot and instruction. Returns actions. Price: $0.04 per call.
- ●POST /v1/ground , map a screenshot + element description to x,y coordinates. Price: $0.03 per call.
- ●POST /v1/parse , free, turns pyautogui code into structured actions.
curl -X POST https://coasty.ai/v1/predict \
-H 'Content-Type: application/json' \
-H 'X-API-Key: $COASTY_API_KEY' \
-d '{
"screenshot": "$(base64 -w 0 screenshot.png)",
"instruction": "Click the submit button.",
"cua_version": "v3"
}'
Replace screenshot.png with a real screenshot file. The API returns JSON with actions and a status. Repeat the loop until status is done.Task runs
- ●POST /v1/runs , server drives an agent to completion. Pass machine_id, task, cua_version, optional instructions, system_prompt, max_steps, deadline_seconds, on_awaiting_human, and webhook_url. Billed $0.05 per agent step.
- ●GET /v1/runs , list runs.
- ●GET /v1/runs/{id} , get a single run.
- ●POST /v1/runs/{id}/cancel , cancel a run.
- ●POST /v1/runs/{id}/resume , resume a run.
- ●GET /v1/runs/{id}/events , stream Server-Sent Events; reconnect with Last-Event-ID.
- ●States: queued, running, awaiting_human, succeeded, failed, cancelled, timed_out.
Task runs are billed $0.05 per agent step, not per API call.
Workflows
- ●POST /v1/workflows , create a versioned JSON DSL of runs.
- ●POST /v1/workflows/{id}/runs , start a run from a workflow.
- ●POST /v1/workflows/runs , ad-hoc inline workflow runs.
- ●Step types: task, assert, if, loop, parallel, human_approval, retry, succeed, fail.
- ●Conditions are structured objects. Variables look like {{inputs.x}} and stepId.field.
- ●Hard guards: budget_cents, max_iterations, deadline_seconds.
- ●Task steps are billed $0.05 each.
Where this beats brittle automation
API-only tools rely on hard selectors that break when UI changes. A computer use agent sees the screen, reads text, and clicks or types like a human. Because it uses vision and actions instead of brittle selectors, it adapts to layout shifts, dynamic content, and new elements without manual selector updates. The Coasty computer use API adds a server-driven agent (task runs) that can pause for human approval, retry, and stream events. You pay only for real steps, not idle loops. Combine this with the MCP server to drive Coasty from Cursor, Claude Desktop, or other MCP clients for a full desktop automation stack.
You now know the exact price and fields for every Coasty computer use API endpoint. Build robust desktop automation with a computer use agent that sees and acts like a human. Get your API key at https://coasty.ai/developers .