OHarness - Your AI Workspace, Free from Model Lock-in
Extension Actions
- Live on Store
Lets the OHarness agent on your computer, running any model you choose, work in the Chrome you're already signed into.
OHarness Browser Bridge lets the OHarness Desktop running on your computer use the Chrome you are already signed into, so it can work on sites that need your login without a second browser or signing in again.
HOW IT WORKS
• The extension talks only to the OHarness app on this computer, through Chrome's native messaging. It has no server of its own and sends nothing over the network.
• It forwards Chrome DevTools Protocol messages between OHarness and the tabs OHarness controls. Every decision about what to do on a page is made by OHarness on your machine.
• Each OHarness conversation opens its own tabs, in a tab group named after the conversation. Tabs you opened yourself are used only when a conversation is told to use one.
• Several conversations, from the desktop app, the web UI or the terminal, can share one browser at the same time.
ALWAYS VISIBLE, EASY TO STOP
• While any tab is being controlled, Chrome shows its own "started debugging this browser" banner. The extension does not hide it.
• The toolbar badge shows how many tabs are being controlled.
• The popup lists every controlled tab, grouped by conversation. "Stop controlling all tabs" hands them all back at once. Cancelling Chrome's banner, or disabling the extension, does the same.
REQUIREMENTS
The extension does nothing on its own. It needs OHarness (https://oharness.ai) on the same computer:
• The OHarness desktop app sets everything up when it starts, or
• With the npm package, run: oharness-browser-host install
Then turn on "Use your own browser" in OHarness settings.
OHarness is free to use with your own model provider key; an optional subscription is available at https://oharness.ai/pricing.
Latest reviews
- Jet Zhang
- This desktop AI agent that isn't locked to one model, and handles the browser smoothly! I've been using OHarness as my everyday desktop agent. What keeps me on it is that it isn't tied to any one vendor. I can switch between Claude, GPT, Gemini, DeepSeek or a local model in the middle of a conversation and the context carries over. I usually plan with a stronger model, then switch to a cheaper one for the repetitive steps, without starting over. This extension fills the last gap. When a task needs the web, the agent uses my own Chrome with the accounts I'm already signed into, so there's no separate browser and no logging in again. I had it go through a few admin dashboards, collect the numbers and write a summary to a local file, while I kept working in my other tabs. Each conversation gets its own tab group and never touches the tabs I'm using. It's also easy to trust. Sessions, memory and keys stay on my computer, and permission rules decide what it can do without asking. Chrome's debugging banner stays visible while it works, and one click in the popup hands every tab back. The only thing I'd add is a per-site allowlist for the browser.