Kahoot AI Solver — Smart Assistant

A Chrome extension that captures the current tab, sends it to a vision-capable AI model, and tells you which Kahoot answer to pick.

Model IDs are not hardcoded. The extension queries the Gemini and OpenAI model endpoints with your own API keys and lets you assign any of the returned models to a keyboard shortcut.

Install

From the Chrome Web Store:

Or load it unpacked:

git clone https://github.com/Ketchio-dev/Kahoot-AI-Solver.git

1. Open chrome://extensions/
2. Enable Developer mode
3. Click Load unpacked and select the cloned folder

Setup

1. Click the extension icon, then the ··· button in the top-right corner to reveal the settings view.
2. Paste a Gemini API key, an OpenAI API key, or both, and click Save Keys & Reload Models.
3. The three dropdowns fill with every model your keys can actually reach, fetched live at that moment.
4. Assign a model to each slot. Your choice is stored locally and used by the matching shortcut.

Get keys from Google AI Studio and the OpenAI platform.

Usage

Shortcut	Action
Alt+Z	Solve with model slot 1
Alt+X	Solve with model slot 2
Alt+C	Solve with model slot 3

You can also press the buttons next to each dropdown in the popup.

While a request is running, a faint dot pulses in the bottom-right corner. When the answer arrives, that dot turns into the shape and color of the correct option for a couple of seconds, and the extension icon flashes the same color:

Color	Kahoot shape	Indicator
Red	Triangle	▲
Blue	Diamond	◆
Yellow	Circle	●
Green	Square	■

Errors surface as a black ! dot in the same corner rather than a popup dialog, with the message in its tooltip and the browser console. The indicator is 14px and click-through, so it never blocks the page.

Holding a shortcut down will not stack requests: while one solve is in flight, further triggers are ignored so you are not billed for duplicate calls.

How live model loading works

models.js calls:

• GET https://generativelanguage.googleapis.com/v1beta/models — paginated, keeping only models that advertise generateContent
• GET https://api.openai.com/v1/models — the response carries no modality metadata, so speech, image-generation, embedding, moderation and legacy completion models are filtered out by id pattern

Results are merged into a single provider-tagged list. If a slot has no saved selection, the first available model is used automatically, so a new install works as soon as a key is saved. Nothing is cached between popup openings — hitting Refresh model list always re-queries the APIs, so newly released models appear without an extension update.

Files

File	Purpose
models.js	Live model discovery for both providers
background.js	Service worker: screenshot capture, model calls, icon feedback
popup.html / popup.js	Settings UI, key storage, slot assignment
content.js / styles.css	Corner indicator for pending, answer and error states
manifest.json	MV3 manifest, permissions, shortcuts

Privacy

API keys and slot selections live in chrome.storage.local on your machine. Screenshots go directly from your browser to the provider you chose. There is no backend server.

Disclaimer

Built for learning and research. Using it to cheat in a graded or competitive setting is on you.