Home » Why Gemini-Config-Checker- Tests Proxy Routes Through Real Signed-In Browser Sessions

Why Gemini-Config-Checker- Tests Proxy Routes Through Real Signed-In Browser Sessions

By

–

Why Gemini-Config-Checker- Tests Proxy Routes Through Real Signed-In Browser Sessions

An open-source tool called Gemini-Config-Checker- is trending on GitHub this week because it solves a specific frustration for people trying to reach Google AI Studio and Gemini through proxy connections. The project, maintained by YellowLaborerMarket, automates the tedious process of finding which proxy configurations actually work.

What The Tool Is And Its Core Purpose

Gemini-Config-Checker- is a chain region scanner. Its job is to test proxy configurations — vless, vmess, trojan, and shadowsocks links — and tell you which ones successfully grant access to Google AI Studio and the Gemini chat interface. Version 0.1.0 is an early pre-release. It works end-to-end on Windows right now; macOS, Linux, and Android builds are provided through the project’s continuous integration pipeline but have not been tested on physical devices yet.

The developer built it because standard connectivity probes give false positives for Google services. Google decides whether an account can access AI Studio or Gemini based on the signed-in user’s region, and it makes that decision after the page has already loaded. A plain HTTP request — even to the Gemini API with a dummy key — returns “fine” from almost anywhere. That means a basic ping or curl check marks every live proxy as healthy, even if Google will ultimately block the exit.

Why Standard Connectivity Probes Fail For Google Services

During development, the author measured exits geolocated to the United States, Hong Kong, and Singapore that were blocked by Google. The same exit IP can flip between allowed and blocked over time. The only reliable signal is what Google actually shows a signed-in user. So the tool opens a real browser, signs in as you, and reads the page the way a person would.

two black flat screen computer monitors
Photo: Fotis Fotopoulos / Unsplash (Unsplash License)

How The Signed-In Browser Check And Advanced Chaining Work

The application has two modes, controlled by a Simple/Advanced switch in the header. In Simple mode, you paste a list of configs — or leave the box empty to use a built-in free list — and the tool tests each one individually. In Advanced mode, you provide a “base config” that already connects reliably to the internet but hits Google’s region wall. The tool then chains free proxy configs behind that base config, searching for an exit that Google accepts.

For the region check, the tool launches a Chrome, Edge, or Brave window in a completely separate browser profile used only by the app. You click “Sign in with Google” once, authenticate normally, and the window closes itself when sign-in is detected. The app never sees your password. The sign-in persists in that isolated profile for future scans. The developer recommends using a secondary Google account because requests arriving from many different IPs can trigger a “suspicious sign-in” notice. A “Remove account from app” button deletes only that private profile; it never touches your everyday browser or accounts.

If you have no browser installed, or you are on Android, you can switch the check type to connectivity-only mode in Settings. That mode skips the browser and the sign-in step, but it cannot verify regional access.

Platform Support And Setup Instructions

On Windows, download the gemini-config-checker-desktop__windows_amd64.zip (or arm64), unzip it anywhere, and run GeminiConfigChecker.exe. WebView2 is built into Windows 10 and 11; if SmartScreen warns you, choose “More info → Run anyway” because the build is unsigned.

On macOS, download the universal zip, unzip it, and move the .app to Applications. On first launch, right-click the app and choose Open. If macOS still refuses, run xattr -dr com.apple.quarantine "/Applications/GeminiConfigChecker.app" in Terminal.

On Linux, install the runtime libraries once — sudo apt install libgtk-3-0 libwebkit2gtk-4.1-0 on Debian or Ubuntu, or your distribution’s equivalents — then extract the tarball and run ./GeminiConfigChecker. If you are on a headless server, use the web server build instead: run ./gemini-config-checker and open http://localhost:8787 in your browser.

On Android, download the arm64 APK, allow “Install unknown apps” for your file manager or browser, and open the APK. A persistent notification reading “Scanner engine is running” is expected; the engine runs as a foreground service so long scans survive the screen turning off. Tap the notification’s Stop action to quit the engine. The Android build runs in connectivity-only mode because Android does not let apps drive a signed-in Chromium, so regional access cannot be verified there.

The web server build serves the same UI on http://localhost:8787 and binds to 127.0.0.1. It rejects foreign Host headers and non-JSON POSTs but has no authentication, so keep it on localhost or tunnel it over SSH instead of exposing it to the network.

Privacy, Security, And Export Features

Traffic to Google is end-to-end TLS. A free-config operator can see the destinations you connect to, but not the content. The project’s README advises against sending sensitive data through free configs. Your configs and settings are stored exclusively in the app’s local storage on your device, and only if you tick the “remember” box. They are never sent anywhere except through your own proxy routes.

Results classify each config as Healthy, Region blocked, or Error. In region mode, Healthy means your “healthy” rule passed; in connectivity mode it only means reachable. Region blocked means Google showed the region wall for at least one product you require, and the AI Studio and Gemini columns show each product with a check or an X. Error covers dead configs, timeouts, or browser problems; hover the label for the reason. Ping shows the latency of one request through the full route.

You can filter by status, search by name or address, sort by ping or name, and select multiple rows for bulk copy. Copy puts the share link on your clipboard for clients like v2rayN, Hiddify, or PattN. “Copy with chain” in Advanced mode or “Copy full config” in Simple mode puts a ready-to-import Xray JSON on the clipboard. The Export tab does the same and can save a .json file, a base64 subscription, or plain links, with a limit on how many entries to include. The exported Xray config uses the classic vnext/servers layout accepted by old and new cores, mixed inbounds on ports 10808 and 10809, pinned DNS, and a UDP/443 block.

For users inside Iran, the tool includes anti-DPI settings tailored for restricted networks, using a custom fragmenting algorithm based on a recipe by patterniha and PattN. Stock xray-core only supports one length and delay per fragment mask, while the recipe uses per-segment values. The app runs the same fragmenting algorithm in a small local forwarder in front of the connection. Only fragment masks are supported; other mask types are rejected with a clear message. The profile should be switched off outside filtered networks because it only adds latency there.

The developer emphasizes that a verdict is a sample, not a guarantee. The exit of a Cloudflare Worker can change between connections, so the same route may pass now and fail a minute later. The recommendation is to re-test the config you intend to keep. The project is licensed MIT; the bundled Xray-core engine uses MPL-2.0 and the fonts use SIL OFL.

Sources

Photo: Nubelson Fernandes / Unsplash (Unsplash License)

Get breaking stories first on our Telegram channel.

More stories by

Most read

AI Tools

Artificial Intelligence

Business

World News