My Real Chrome v2 (publishable — scripting + native messaging, no debugger)
The clean-room, Web-Store-approvable rebuild of Omniscio's Chrome bridge, driving your real signed-in Chrome without the debugger permission — how to enable it, what it deliberately cannot do, and why driving it does not leave you anywhere else — the one switch a drive makes is forced by Chrome, lasts for the length of the command, and puts your own tab back.
On this page
What it is
"My Real Chrome v2" is the user-facing name; the code namespace is real-chrome-bridge-v2.
Status: in development — hidden by default, desktop-only. Reveal via Settings → Features → "My Real Chrome v2" or AMC_SHOW_REAL_CHROME_BRIDGE_V2=1. Ships OFF; nothing runs until you enable it, load the companion extension, and grant per-site access.
A clean-room, Chrome-Web-Store-approvable rebuild of My Real Chrome — Full (deleted on 2026-10-17) that drives your real, signed-in Chrome without the debugger permission. It uses chrome.scripting (inject a driver into the page), chrome.tabs.captureVisibleTab (screenshots), and native messaging (transport) — the same architecture as Anthropic's published "Claude" Chrome extension. It was built separately from v1, and it is the Chrome extension Agent Web's my-Chrome engine drives.
Where to find it
Nothing appears until you turn it on. The feature is revealed by flipping Settings → Features → "My Real Chrome v2" (it ships hidden and off). From then on the surface lives in Chrome rather than in Omniscio: you load the companion real-chrome-extension-v2/ folder as an unpacked extension, and per-site access is granted by clicking the extension icon and pressing Allow this site (or Allow all sites). Driving it is an agent's job, through Agent Web (amc-web --engine my-chrome, see Agent Web, part 2) — there is no bridge screen inside the app.
How it behaves
What the bridge will tell you it is allowed to do
GET /real-chrome-v2/status reports the host access the paired extension holds: the origin patterns
Chrome currently grants it, and whether captureVisibleTab will work at all — which it will only
under the whole-origin Allow all sites grant. The extension re-reports on every grant change, so
the answer follows a grant or a revoke without anyone asking, and it goes back to null when nothing
is connected. null means nobody has said; it is deliberately NOT the same as "no sites allowed", so
an empty grant and an unreported one can never be confused.
Why it exists (vs. v1)
- Publishable. The Web Store rarely approves consumer extensions that use
debuggerto drive the browser; v1 therefore shipped load-it-yourself only, and has since been deleted. v2's permission set (scripting,tabs,webNavigation,storage,alarms,nativeMessaging,cookies+ optional per-site host access) is the approvable shape.cookiesexists for one job — signing the browser out of claude.ai before an automated account sign-in — and the extension confines it to that site with a fixed allow-list, since Chrome offers no narrower grant for the API. - Not poisoned by other extensions. v1's debugger session was poisoned (CDP -32000) the instant a page contains another extension's frame (LastPass, Superhuman). v2 uses no debugger, so it has no such wall.
The one command that touches cookies
clear-cookies removes a site's cookies so an automated sign-in can start from a signed-out browser.
It is the only command that acts on cookies and it is deliberately narrow:
- One site, fixed in the extension. The host allow-list is a constant in the extension itself — a
caller names a host and anything outside it is refused by name before a single cookie is read. The
manifest cannot enforce that, because Chrome gates
chrome.cookieson the host permissions the extension HOLDS, and it holds the optional whole-site grant; the constant is the real limit. - The clearance survives. Cloudflare's
cf_clearance,__cf_bmand_cfuvidare left in place. Removing them makes the next request answer a challenge page, which reads to the automation as an unreadable session rather than a signed-out one. - It still asks permission the ordinary way. The command passes the same user-granted site check every drive passes, so a site the user has not granted is refused rather than cleared.
The AI-control banner
Every tab the extension drives shows a loud "Omniscio AI is controlling this tab" strip across the top of the page, with the Omniscio orb and a matching border, and the tab's own title is marked so it is identifiable from the tab strip. The strip takes its own space, so it pushes the page down rather than covering the first rows of it.
On a tab a session is driving, the strip names that session, and the name is a link: click it and Omniscio opens on the session, so you can go straight from the page you are watching to whoever is working in it. The name is the only clickable part of the strip — everything else stays transparent to the page underneath — and a name too long for the strip is shortened with an ellipsis rather than running off the page. Every drive runs on the machine-wide key, the only one the routes take, and the name still comes through because the app tells the extension which session, and which name, each command was made for. The link is only ever built from a validated session id, the name is only ever shown beside the id it arrived with, and a tab you drive yourself with the machine-wide key (you, at the CLI, with no session behind it) shows the strip as plain text with no name and no link at all.
What it is NOT
- Not your existing tabs — by default. The extension drives ONLY tabs it opened (the fresh-tab guardrail). The default-off "drive any tab" opt-in is the one deliberate exception.
- Not the tabs you open next to it. The default-off popup setting follows only a popup that a PAGE in a driven tab opened, such as a "Sign in with Google" window. A tab you open yourself — Ctrl+T, the + button, an address typed with Alt+Enter, Duplicate — is never taken over, even while a driven tab is the one in front.
- Not on the network. Transport is Chrome's native messaging to an app-registered host, relayed to the app over a private local pipe — no TCP port at all.
- Not reachable from your phone. Desktop-only: gated CLI routes, and the status channel is on the mobile/web bridge blocklist.
Setup
- Enable — Settings → Features → "My Real Chrome v2" (or
AMC_SHOW_REAL_CHROME_BRIDGE_V2=1). The app registers the native-messaging host with Chrome automatically: an HKCU registry entry on Windows; on macOS and Linux a host manifest in each installed browser'sNativeMessagingHostsfolder (Chrome and Edge on macOS; Chrome, Chromium and Edge on Linux) that points at a smallshlauncher. A browser that is not installed is skipped — its folders are never created. - Load the extension —
real-chrome-extension-v2/as an unpacked extension (chrome://extensions → Developer mode → Load unpacked). There is no token to paste — it connects automatically. - Grant access — click the extension icon and press Allow this site (or Allow all sites). Chrome
requires this click, so an agent can never silently gain access to a new site. Revoke any time in Chrome.
The popup tells you what you hold. Under the connection banner it lists the sites currently granted
and says in plain words whether screenshots are on, so the button you pressed and the access it bought
are never out of step — and a per-site grant answers by naming what it covers ("reading and clicking
only; screenshots need Allow all sites") instead of a bare "access granted".
Screenshots need "Allow all sites": Chrome gates
captureVisibleTabon the extension holding the whole-origin<all_urls>permission, and a per-site grant cannot satisfy it. On a per-site grant alone a page is still readable, clickable and typable — just not photographable.
Known limits (scripting can't match a debugger)
These are documented, expected limits — not bugs — of the no-debugger architecture:
- Trusted input: synthetic events (
isTrusted:false); a few sites that gate on genuine user events won't respond. - File uploads: not supported (no debugger file-set); a file picker needs the user.
- Screenshots: visible area of the active tab only, and only under the whole-origin Allow all
sites grant — Chrome gates
captureVisibleTabon the extension holding<all_urls>, so with a per-site grant alone the capture is refused by name. A background tab whose window is NOT the one in front is briefly activated then restored (nobody sees it); a background tab in the window the user IS looking at is refused rather than shown, because the switch would take over their screen — see Background-first below. - Read-page: a DOM-derived accessibility approximation, not the browser's real accessibility tree.
- Not captured/controlled: native JS dialogs, console messages, network response bodies, device/network emulation.
When it says it isn't connected, or has no access to this site
The two refusals people actually hit, and what each one means:
| What you see | What it means | What to do |
|---|---|---|
| The popup says it isn't connected to Omniscio | The extension is loaded, but Chrome could not reach the app's native-messaging host — the app is closed, the feature is off, or Chrome started before the host was registered | Keep Omniscio open and confirm Settings → Features → "My Real Chrome v2" is on. Then close Chrome completely (every window) and reopen it — Chrome reads the host registration at launch, so a host registered after it started is invisible until a restart. Still disconnected after that? Toggle the feature off and on, reopen Chrome, and check the popup again |
| The agent (or the popup) says it has no access to this site | The extension starts with no site access by design; nobody — including an agent — can grant it for you | Click the extension icon and press Allow this site (or Allow all sites). Grants are per site, so a new domain needs its own click. chrome:// pages, the Chrome Web Store and other browser-internal pages can never be granted |
| Screenshots are refused, but reading and clicking work | A per-site grant never satisfies Chrome's whole-origin requirement for captureVisibleTab |
Press Allow all sites in the popup. A per-site grant still reads, clicks and types — it just can't photograph the tab |
| The agent says no tab was found, or drives the wrong one | The bridge drives tabs you already have open; it does not open pages on its own authority | Open the page in Chrome first, then let the agent read or act on it |
If none of that lands, the fastest route is an agent in any session — it can read the bridge's own logs and say which half refused. Attach Settings → Diagnostics → Export diagnostic bundle when you send feedback with the toolbar Feedback button, and say which of the rows above you hit.
Background-first — it does not STAY on the user's screen
Opening is invisible, and driving never leaves the user anywhere else. This is deliberate: the user was clear that the bridge must "work in the background unless the agent specifically pulls the screen to the front."
- Opening a tab puts it
active:falsein the user's own last-focused window. It never mints a focused window, and never switches them off what they are reading. - Reading and acting DO move the tab to the front for the length of the command, and put the user's
own tab back about a second after the agent's last command — so the many commands of one task (read,
check, act, wait for the page to settle, read again) switch the screen once, not once each. A
switch-taborforegroundin that second is kept, never undone. The rule is runs script in the page:read-page,get-page-text,click,hover,type,press-key,select-option,fill-form,navigate-backandwait-for.navigateis NOT one of them — it retargets a tab, which leaves a background tab in the background — and neither isreload. Chrome will not run a script in a tab the user is not looking at, so there is no way to drive a page without this. It is roughly a third of a second for a read or a click, but await-forholds the front for its WHOLE wait (up to its 60s limit). If the tab is already the one on screen nothing changes on screen at all, and if the user switches tabs during a drive their choice wins — the tab is put back only while the driven tab is still the one in front. screenshotis the one read that REFUSES to move the screen —captureVisibleTabphotographs a window's VISIBLE tab, so shooting a background tab means bringing it forward first. It only does that when the switch is invisible (the tab is already on screen, or its window is behind something). Otherwise it refuses, namingforegroundas the way to do it on purpose.foregroundis the deliberate screen-grab, and the only op that raises the Chrome window. Reach for it when a HUMAN needs to see the tab — a login hand-off, a "look at this" moment.
For a login flow, use the drive → hand off → resume pattern: drive to the sign-in, foreground the
tab for the user to finish, then resume on the now-authenticated tab. foreground activates the tab and raises
the Chrome window itself (TOPMOST → NOTOPMOST, restoring it first if minimized) — the extension's
chrome.windows.update{focused} alone cannot, because Windows silently vetoes Chrome's SetForegroundWindow
when Chrome is not already the foreground process. So the hand-off works even when Chrome is behind another
window or minimized; never tell the user to un-cover or un-minimize it by hand.
Related
real-chrome-bridge.md records v1, the debugger-based original this rebuild replaced, deleted on 2026-10-17. Agents reach this bridge through Agent Web (part 2, "Your own everyday Chrome"). For the other ways Omniscio drives a browser, see embedded-browser.md and browser-logins.md. The invariants this bridge must hold, and the map of tests that guard them, are locked in real-chrome-bridge-v2-contract.md.
Last verified 2026-11-11