Cloud-hosted AI browser agent that executes multi-step web tasks autonomously. Give it natural language instructions and it navigates, clicks, fills forms, and extracts data from any website using a real browser with screenshot-based vision. Supports multiple LLMs including Browser Use 2.0, Claude, GPT-4, and Gemini.
View documentationUse Browser Use when your agent needs to interact with websites that don't have APIs — filling out forms, scraping dynamic content, navigating multi-step workflows, or automating repetitive web tasks. Ideal for data collection agents, form submission bots, web testing, and any workflow that requires a real browser with JavaScript rendering.
Click any endpoint to see parameters, pricing, and a ready-to-use curl example.
Start an automated browser task. Provide natural language instructions and the agent navigates websites, clicks elements, fills forms, and returns results. Supports configurable LLM, max steps, start URL, and structured output schemas.
/services/browser-use/run-task| Name | Type | Required | Description |
|---|---|---|---|
| task | string | Yes | Natural language instructions for the browser agent to execute (1-50,000 chars) |
| maxSteps | number | No | Maximum number of agent steps before stopping (default: 100, max: 10,000) |
| startUrl | string | No | URL to navigate to before starting the task |
| llm | string | No | LLM model — browser-use-2.0 (default), gpt-4.1, claude-sonnet-4-5-20250929, gemini-2.5-pro, o3, etc. |
| structuredOutput | string | No | Stringified JSON Schema to enforce structured output format |
curl -X POST \
"https://api.indiegent.com/services/browser-use/run-task" \
-H "Authorization: Bearer <your-token>" \
-H "X-Agent-Wallet: my-agent" \
-H "Content-Type: application/json" \
-d '{
"task": "Go to hackernews.com and get the titles of the top 5 posts",
"maxSteps": 50
}'api.indiegent.com/services/browser-useUSDC · Automatic
Payment protocols handled by IndieGent
Tempo