Camel AI
An AI that operates real software through its screen — tests any
web or desktop app and hands a fix-brief to your AI coder, or automates your
workflows. No target-app API needed.
Windows
macOS
Linux
irm https://raw.githubusercontent.com/DilawarShafiq/camel-ai/main/install.ps1 | iex
Copy
Paste that into your terminal. It installs everything and opens a setup
wizard — pick any LLM (Gemini's free tier by default) and paste one key.
Watch it log in by itself. Then just tell it what to do:
1. Run the one line above (or download the installer).
2. The wizard connects your AI “brain” — Gemini (free), OpenAI, Claude, OpenRouter, Groq, local models, or any custom endpoint.
3. Type a goal: “audit every button on my site” or “log into the portal and download my invoices.” Do any 2FA yourself when it pauses.
Everyday commands
camel dashboard — open the app (audit, tasks, schedule, see windows)
camel audit <url> — test a site → HTML report + a fix-brief for your AI coder
camel login <url> then camel audit <url> --real-browser — test behind a login
camel run "<goal>" — do a task in plain English (pauses for 2FA)
camel jobs add "<url>" --at 09:00 + camel daemon — run on a schedule
pip install aicamel or docker run -p 8765:8765 ghcr.io/dilawarshafiq/camel-ai — other installs
🧪 Testing → fix brief Finds dead buttons, JS errors, broken requests, and accessibility issues, then emits a ranked, machine-readable brief your AI coder can act on.
🤖 Automation, any industry Medical billing, financial, marketing — anything with a UI, in a browser on any OS.
🖥️ Three drivers Web (Playwright), native desktop (Windows/macOS/Linux accessibility), and a universal vision fallback.
🔐 You keep control It pauses for 2FA/CAPTCHA/login so you finish sensitive steps — it never handles your codes.
Open source · GitHub
The only API camel ever uses is an LLM API (or your subscription) — never the target app's.