Banquish is a Mac app that turns any website into a workspace. Pull the pieces you need onto one canvas, keep them live, and let an agent run the errands.
I'm Gunn, the maker of Banquish.
This started with a frustration I couldn't shake: the web is the most powerful tool we have, but we experience it through a stack of tabs and windows that someone else decided the shape of. You can't take the one part of a page you care about and put it next to the one part of another. You just tab back and forth, all day.
Banquish lets you clip living, interactive pieces of the web and arrange them on your own canvas. Not screenshots — the real, live pages, still scrollable and clickable. Pull the tracking widget, the doc, the dashboard, and the chat you actually use into one space laid out the way you think, not the way each site wants.
And there's an AI agent built in that handles the tedious parts — the repetitive clicking, gathering, and watching — so you stay on the judgment calls.
I built it around one idea: deconstruct the order the web imposes, keep only what's essential, and build your own. It's a native Mac app, and the core is free — you only pay when you lean on the agent.
I'd genuinely love your honest reactions, especially:
- What's the first workspace you'd build?
- Where does the live-panel idea click — or not?
I'll be here all day answering everything. Thank you for taking a look 🙏
Report
What I'd build first is a small support dashboard, pulling the Tellacity Search Console panel, the live review feed, and a doc of open bugs onto one canvas, since right now that's three separate tabs I flip between constantly and lose context each time. The live-panel idea clicks for me specifically because those pieces update on their own timelines, a review can land any moment, so a static clip would go stale within the hour.
Where I'd want more detail is on the AI agent handling the tedious clicking and gathering across live pages that are still fully interactive. If the agent is clicking through a real, live site on my behalf, does it operate inside an isolated context so a site can't tell the difference between the agent and me manually clicking, and is there any rate limiting so it doesn't end up looking like automated traffic to a service that might flag or block it.
@thys_beesman Great questions — this is the part I'm most careful about.
The key design decision: the agent acts as you, inside your own session, not as a separate bot. It's not a headless browser or a cloud automation context — it's the same live WebView you're using and watching, so it carries your real cookies, your login, your device fingerprint, and your own IP. There's no fresh datacenter profile, which is usually the thing that gets flagged as "automated traffic" in the first place.
Report
Congrats to the launch! Rebuilding the web interface around personal workflows instead of forced tab structures is a huge advantage. Honestly, I can see this easily replacing half of my daily open browser windows.
Quick question on the workspace layout dynamics: Can you group or link interactions between different live panels (e.g., clicking a link or triggering an action in Panel A automatically updates or context-syncs Panel B), or is each panel strictly running in its own isolated WebView for now?
@hannesh Thank you — "replacing half your open windows" is exactly the outcome I'm going for
On the layout dynamics: both are true, by design. Each panel is its own live WebView with its own session (so a logged-in dashboard and a public page stay properly isolated) — but they're not siloed. You can link them with Flows: pull a value out of Panel A, and use it to drive Panel B — navigate it to a new URL built from that value, or run a click/type action in it. So "grab the order ID from this panel → search for it in that one," or "take the address here → route it over there," becomes one repeatable step instead of manual copy-paste between tabs.
Report
@geonhyeok_boo Awesome, thanks for the clear explanation! Linking live panels via Flows sounds super powerful. Looking forward to testing it out!
Report
"living pieces of the web" is the right way to describe this - most clipping tools just take a screenshot and the source page might as well not exist anymore. does a clipped piece keep updating live if the original page changes, or is it more of a snapshot that you'd have to manually re-clip later to refresh?
@omri_ben_shoham1 Exactly that's the whole reason I built it this way. A Banquish clip isn't a screenshot. it's the real, live page, just masked down to the piece you picked. The scripts, event handlers, and your logged-in session all stay intact and running.
So if the source updates — a live dashboard, a feed, a stock ticker, a doc someone else is editing — your clipped panel updates right along with it, in real time. You never re-clip to refresh. It's still fully interactive too: scroll it, click into it, type in it.
The one time you'd re-pick is if the site itself gets restructured (a redesign that moves or removes the element you anchored to) — Banquish flags that with a staleness badge and helps you re-anchor in a couple of clicks. But that's a rare structural change, not something you do to keep content fresh.
Report
the "same live session as you, not a separate bot" design is clever for the automated-traffic problem, but it cuts the other way too - if the agent misreads a page and clicks the wrong thing, it's using your real logged-in session to do it. is there a confirm step before anything irreversible (a purchase, a send, a delete), or does "let an agent run the errands" mean it can go all the way through on its own once you kick it off?
@galdayanThat’s exactly the right concern. The agent acts in the same live session as you, so we don’t treat safety as something the model should “figure out” on its own.
Banquish has deterministic gates before actions: targets are classified, and sensitive actions like form submissions, password fields, payment fields, purchases, sends, and deletes can require an explicit confirmation before the agent continues.
today that confirmation is a setting, and it isn't on by default
Report
@geonhyeok_boo appreciate the honesty on that last part. genuinely surprised it's opt-in rather than opt-out given the framing though - if the whole pitch is the agent uses your real session so it doesn't look automated, the moment it's wrong on a purchase or a delete the blast radius is on you, not a sandboxed bot. seems like the safer default would be confirmation on, with power users opting out once they trust it, not the other way around
@galdayan That’s a fair push, and I appreciate you calling it out.
The reason I made it configurable was that I was worried too many confirmations would make the agent feel annoying or slow in everyday use. But I think you’re right about the default: if the agent is acting in your real logged-in session, the safer baseline should probably be confirmation on for irreversible actions, with power users opting out intentionally.
I’m going to seriously revisit that default. Thanks for pressure-testing this.
Report
@geonhyeok_boo good to hear, and honestly the fact that you're open to flipping it says more than the current default did. small thing worth considering when you do the switch: maybe scope which actions get the opt-out first (say, anything involving money or deletes stays hard-confirmed even for power users) rather than one global toggle, so someone can't accidentally disable it once and forget.
Report
Where's the support section? I am having issues getting this to work. Ty
@andrea_jSorry you're hitting a snag, and thank you for flagging it! I'll get you sorted.
The fastest way is to email me directly at support@banquish.space — I read these myself and reply quickly.
To help me pin it down, whenever you get a sec: what happens exactly (a screenshot or the error is perfect), and roughly what step you're on — installing, first launch, capturing your first panel? i'll take it from there
Local: it already is. Your spaces, panels, and layouts are stored locally on your Mac — the only thing that touches the cloud is your account/billing and the AI agent relay. Your workspace itself never leaves your machine.
Sync: not yet, but it's my obvious next step. Because a workspace is already a self-contained local document, syncing it across your own devices (ideally end-to-end encrypted, so I never see the contents) is very much the direction I want to go.
Mac + Android: Mac is home first. I'll be straight with you — the rendering engine is built deep on Apple's WebKit, so Android isn't a quick port
Banquish — Make the web your workspace
Banquish — Make the web your workspace
@thys_beesman Great questions — this is the part I'm most careful about.
The key design decision: the agent acts as you, inside your own session, not as a separate bot. It's not a headless browser or a cloud automation context — it's the same live WebView you're using and watching, so it carries your real cookies, your login, your device fingerprint, and your own IP. There's no fresh datacenter profile, which is usually the thing that gets flagged as "automated traffic" in the first place.
Congrats to the launch! Rebuilding the web interface around personal workflows instead of forced tab structures is a huge advantage. Honestly, I can see this easily replacing half of my daily open browser windows.
Quick question on the workspace layout dynamics: Can you group or link interactions between different live panels (e.g., clicking a link or triggering an action in Panel A automatically updates or context-syncs Panel B), or is each panel strictly running in its own isolated WebView for now?
Banquish — Make the web your workspace
@hannesh Thank you — "replacing half your open windows" is exactly the outcome I'm going for
On the layout dynamics: both are true, by design. Each panel is its own live WebView with its own session (so a logged-in dashboard and a public page stay properly isolated) — but they're not siloed. You can link them with Flows: pull a value out of Panel A, and use it to drive Panel B — navigate it to a new URL built from that value, or run a click/type action in it. So "grab the order ID from this panel → search for it in that one," or "take the address here → route it over there," becomes one repeatable step instead of manual copy-paste between tabs.
"living pieces of the web" is the right way to describe this - most clipping tools just take a screenshot and the source page might as well not exist anymore. does a clipped piece keep updating live if the original page changes, or is it more of a snapshot that you'd have to manually re-clip later to refresh?
Banquish — Make the web your workspace
@omri_ben_shoham1 Exactly that's the whole reason I built it this way. A Banquish clip isn't a screenshot. it's the real, live page, just masked down to the piece you picked. The scripts, event handlers, and your logged-in session all stay intact and running.
So if the source updates — a live dashboard, a feed, a stock ticker, a doc someone else is editing — your clipped panel updates right along with it, in real time. You never re-clip to refresh. It's still fully interactive too: scroll it, click into it, type in it.
The one time you'd re-pick is if the site itself gets restructured (a redesign that moves or removes the element you anchored to) — Banquish flags that with a staleness badge and helps you re-anchor in a couple of clicks. But that's a rare structural change, not something you do to keep content fresh.
the "same live session as you, not a separate bot" design is clever for the automated-traffic problem, but it cuts the other way too - if the agent misreads a page and clicks the wrong thing, it's using your real logged-in session to do it. is there a confirm step before anything irreversible (a purchase, a send, a delete), or does "let an agent run the errands" mean it can go all the way through on its own once you kick it off?
Banquish — Make the web your workspace
@galdayanThat’s exactly the right concern. The agent acts in the same live session as you, so we don’t treat safety as something the model should “figure out” on its own.
Banquish has deterministic gates before actions: targets are classified, and sensitive actions like form submissions, password fields, payment fields, purchases, sends, and deletes can require an explicit confirmation before the agent continues.
today that confirmation is a setting, and it isn't on by default
@geonhyeok_boo appreciate the honesty on that last part. genuinely surprised it's opt-in rather than opt-out given the framing though - if the whole pitch is the agent uses your real session so it doesn't look automated, the moment it's wrong on a purchase or a delete the blast radius is on you, not a sandboxed bot. seems like the safer default would be confirmation on, with power users opting out once they trust it, not the other way around
Banquish — Make the web your workspace
@galdayan That’s a fair push, and I appreciate you calling it out.
The reason I made it configurable was that I was worried too many confirmations would make the agent feel annoying or slow in everyday use. But I think you’re right about the default: if the agent is acting in your real logged-in session, the safer baseline should probably be confirmation on for irreversible actions, with power users opting out intentionally.
I’m going to seriously revisit that default. Thanks for pressure-testing this.
@geonhyeok_boo good to hear, and honestly the fact that you're open to flipping it says more than the current default did. small thing worth considering when you do the switch: maybe scope which actions get the opt-out first (say, anything involving money or deletes stays hard-confirmed even for power users) rather than one global toggle, so someone can't accidentally disable it once and forget.
Where's the support section? I am having issues getting this to work. Ty
Banquish — Make the web your workspace
@andrea_jSorry you're hitting a snag, and thank you for flagging it! I'll get you sorted.
The fastest way is to email me directly at support@banquish.space — I read these myself and reply quickly.
To help me pin it down, whenever you get a sec: what happens exactly (a screenshot or the error is perfect), and roughly what step you're on — installing, first launch, capturing your first panel? i'll take it from there
@geonhyeok_boo Thanks - I'll pop an email over
Love this kind of app. Will there ever be a way to synch somehow or make local? On mac and android?
Banquish — Make the web your workspace
@andrea_jThank you 🙏 Two honest answers:
Local: it already is. Your spaces, panels, and layouts are stored locally on your Mac — the only thing that touches the cloud is your account/billing and the AI agent relay. Your workspace itself never leaves your machine.
Sync: not yet, but it's my obvious next step. Because a workspace is already a self-contained local document, syncing it across your own devices (ideally end-to-end encrypted, so I never see the contents) is very much the direction I want to go.
Mac + Android: Mac is home first. I'll be straight with you — the rendering engine is built deep on Apple's WebKit, so Android isn't a quick port