Give Arlo a task in plain language. It reads the tab you’re looking at, opens tabs for you, and streams its answer as it works. Your model, your key, no backend.
A small, focused assistant that works on the page you’re on.
Ask about the page in front of you. Arlo reads it at the level of detail the task needs — compact, detailed, or raw HTML.
No new tab, no context switch. Arlo sits next to the page and streams its answer as it works.
Every tab Arlo opens goes into an “Arlo” tab group, so you can see them all and close them in one click.
One click halts a running task. You always see what Arlo is doing while it does it.
Point Arlo at any OpenAI-compatible endpoint — OpenAI, Anthropic, a proxy, or a local model server. Keep several profiles.
Nothing is granted at install. Site access is requested per origin, only when needed, and you can revoke it in Chrome.
There’s no server to start — an endpoint and a key are the whole setup.
Click the Arlo toolbar icon to open the side panel.
In the options page, enter an OpenAI-compatible endpoint and your API key. Choose whether the key is remembered or kept for the browser session only.
Chrome asks for access to your model’s site, and only if you choose, to all pages. Both are optional and revocable.
Describe what you want. Watch Arlo work, and stop it whenever you like.