Multi-agent AI,
built for developers.
Eight specialized AI engines — research, code generation, analysis, live web search, and browser computer use — unified under one precise interface.
How each engine is built
Universal Chat
Streaming chat over the Gemini API with markdown + code rendering.
Uses: Gemini 2.5 Flash
Best for: Quick questions, drafting, explanations.
Agent Console
A raw agent loop you can watch think and act step by step.
Uses: Gemini 2.5 Flash
Best for: Debugging agent behaviour, power users.
Atlas Agent
Research + reasoning agent with live web search.
Uses: Gemini + Google Search
Best for: Fact-checked answers, current info.
Multi-Agent
A crew — planner, researcher, executor, critic — that splits a task.
Uses: Gemini pipeline
Best for: Compound tasks with several steps.
Nova Agent
4-agent pipeline that opens a real browser, searches, extracts, critiques.
Uses: Gemini + Playwright
Best for: Deep research with real sources.
Atlas Pro
Multi-step search + live browser tasks + Google Docs writing.
Uses: Gemini + Playwright
Best for: Reports, live web tasks, documents.
Browser Search
Fetches real pages, strips them to clean text, answers with citations.
Uses: Gemini + live fetch
Best for: Answers grounded in specific pages.
Computer Use
Vision loop that sees the screen and drives the mouse & keyboard.
Uses: Gemini vision + native
Best for: Doing tasks inside any app.
How to use it
Pick an engine
Choose from the sidebar based on your task — chat for questions, Nova/Atlas Pro for research, Computer Use to control apps.
Type your goal
Describe what you want in plain words. Toggles like Live typing or Computer Use switch how it runs.
Watch it work
Agents show their steps live — searches, pages read, actions taken — so you can follow along.
Get the result
A cited answer, a written doc, or a task done on your screen. It stops before anything irreversible.
Nexus AI vs. a typical AI agent
Frequently asked
Is it free?
Yes — it runs on the Gemini free tier with your own key. There's a daily quota that resets each day; add more keys or a paid key for higher limits.
Does it really control my computer?
Computer Use can, via a vision loop that moves the mouse and types. It always asks before anything irreversible like buying or sending.
Where does my data go?
Your keys and history stay on your machine. Only the specific text/screens a task needs are sent to the model you chose.
Can I use Claude instead of Gemini?
Yes, if you add an Anthropic API key — the app switches to Claude automatically. Otherwise it uses Gemini.
What makes it different from a normal chatbot?
It doesn't just talk — it searches, opens real browsers, controls apps, and remembers what went wrong so it doesn't repeat mistakes.