Hanzo Extension
Your browser, as a tool an agent can use.
It opens tabs, reads pages, and clicks things — in the session you are already signed into.
How it works
The extension registers as a provider on the local fabric. Whatever can reach the fabric can then reach a tab.
It uses your session
The agent drives the browser you are already signed into, so a page behind a login is a page it can read. Nothing has to be scraped from the outside.
No port to open
The extension talks to the local Hanzo daemon through a native messaging host over a Unix socket. There is no WebSocket, no localhost port, and nothing else on the machine can dial in.
Connected means registered
It only reports itself connected once the daemon lists it as a provider. An open pipe to a wedged process is the failure this replaced, and it is checked on every tick.
Three browsers, one behaviour
Chrome, Firefox and Safari builds come out of the same source. The evaluate rule is shared, so the same snippet from a caller behaves identically in all three.
Tabs are addressable
Each browser registers separately, so listing tabs asks the browser that is active — and naming a tab id from another browser's list is how you reach the other one.
The rest of your tools, too
The same repo carries connectors for the software your work already lives in — documents, tickets, design files, mail, calendars — so an agent reaches them the same way it reaches a tab.
Read a page as text
It hands back page content, console messages and network requests, which is what makes it useful for checking your own site rather than just browsing someone else's.
Installs from a file
Load it unpacked in Chrome or as a temporary add-on in Firefox while you are working on it. The build produces all three targets in one command.
Available on All Major Browsers
Install the Hanzo Extension on your preferred browser
Chrome
Perfect integration with Google Chrome browser with all features supported.
Firefox
Fully compatible with Mozilla Firefox for privacy-conscious users.
Edge
Enhanced experience on Microsoft Edge with optimized performance.
More Browsers
Safari, Opera, and other Chromium-based browsers are also supported.
What you would use it for
Mostly the work that only exists inside a browser window
Check the change you just made
An agent that edited a page can open it, look at what rendered, and read the console — instead of telling you it thinks the fix worked.
Behind a login
An internal dashboard, a vendor console, an admin panel. It runs in the session you already have, so there is no service account to provision for a one-off task.
Debug the request, not the guess
Network requests and console messages come back as data. When a page misbehaves, the agent can read what the browser saw.
The long boring form
Multi-step flows that exist only in a UI and have no API behind them. It fills them in and tells you where it got stuck.
Install it and open a tab
Once it is registered, your coding agent can check the page it just changed instead of asking you to.