browser-use
Browser-controlling agent. Fills forms, navigates pages, scrapes data autonomously. 🌐
Upstream: browser-use/web-ui ↗
Resources
| vCPUs | 2 |
|---|---|
| Memory | 4096 MB |
| Volume | 10 GB |
| Workspace UI | Web (iframe) |
Create one
gentity compute create --agent browser-use --model claude-sonnet-4-5First run
The workspace URL is the browser-use web UI on port 7788. Inside the container is a full Chromium + Playwright stack plus a supervisord process tree. Click in the UI to give the agent a goal; it drives a headless browser to complete it.
The image also exposes VNC and a Chrome debugger internally, but Gentity only routes the web UI externally — the others stay accessible only over Fly's private network.
Models
| Provider | Name | Label |
|---|---|---|
anthropic | claude-sonnet-4-5 | Claude Sonnet 4.5 (recommended) |
openai | gpt-4o | GPT-4o |
anthropic | claude-haiku-4-5 | Claude Haiku 4.5 (fast) |
google | gemini-3-pro | Gemini 3 Pro |
deepseek | deepseek-chat | DeepSeek Chat |
openrouter | openrouter/auto | OpenRouter (100+ models) |
Notes / known limits
This is the heaviest agent in the catalog (4GB RAM, 2 vCPU, 10GB volume). First boot pulls Chromium so the spin-up takes ~20s rather than the usual 8s.