An AI that actually uses your browser.

Tell it what you want done. It reads pages, clicks buttons, fills forms, writes code, plays games, and finishes the job. No manual steps. No copy-paste. It just does it.

See it in action
Tensor — Homework Helper live recording

Tensor reads the quiz, researches answers, fills in responses, and submits — autonomously.

Wordle live recording

Reads the board, tracks constraints, solves the puzzle.

// What Tensor Does
>_

Full Browser Control

Tensor reads pages, clicks buttons, fills inputs, navigates links, scrolls, and types. It uses the real DOM, not screenshots. It works on any website.

Form Filling

Give it a form and it fills every field. Job applications, checkout flows, surveys, multi-page wizards. It reads the fields, figures out what goes where, and fills them.

Deep Research

Ask a question and it searches the web, visits multiple sources, reads them, cross-references facts, and delivers a synthesis with citations. Not a summary of one page.

Play Any Game

Wordle, chess, puzzles, trivia. Tensor reads the game state from the DOM, tracks constraints, learns from feedback, and plays optimally. It figures out the rules on its own.

Write & Edit Code

Full terminal access. It runs npm install, writes files, starts servers, tests code, and fixes errors. It can build entire projects from a description.

Multi-Tab Agents

Deploy multiple AI agents across browser tabs simultaneously. Each one researches a different source, then results are synthesized into one comprehensive answer.

// How It Works

Reads the Page

Tensor extracts the full DOM tree and understands every element on the page. It knows what's a button, what's an input, what's a link. It uses element references, not brittle coordinates.

Thinks, Then Acts

After every action, Tensor analyzes the result. What changed? Did it work? It tracks constraints, learns from errors, and never repeats the same mistake.

Adapts to Any Task

50 expert skills auto-activate based on what you ask. Shopping triggers price comparison mode. Coding triggers terminal mode. Games trigger constraint tracking. No manual setup.

Vision When Needed

For visual content like charts, canvas games, or PDFs, Tensor takes a screenshot and uses vision AI to understand what it sees. DOM for structure, vision for visuals.

Local Computer Access

Full terminal and filesystem access through a local relay. Install packages, run scripts, read and write files, manage git repos, start dev servers. All from natural language.

Web Search Built In

Tensor searches the web natively. No need to open Google. It finds real URLs, navigates to them, reads the content, and brings back specific facts.

// Real Use Cases

"Fill out this job application"

Tensor reads every field, fills your name, experience, answers screening questions, and submits. Multi-page forms, dropdowns, date pickers, all handled.

"Compare these 5 products"

Visits each product page, extracts price, specs, reviews, shipping info, and builds a comparison. Crosses multiple sites for the best deal.

"Build me a todo app"

Creates the project, installs dependencies, writes the code, starts the dev server, and opens it in a new tab. Full-stack development from a sentence.

"Research AI regulation in the EU"

Deploys 3 agents to different sources simultaneously. One reads official documents, one checks news, one finds expert analysis. Results merged into a full report with sources.

Agent Swarm

Deploy multiple AI agents across browser tabs simultaneously. Each agent gets a detailed brief, navigates independently, and reports back. Results are auto-synthesized into one answer.

Agent Nova
Official Sources
Reading government documents and official reports for verified data and policy details
Agent Flux
News Coverage
Scanning major news outlets for recent developments, expert quotes, and public impact
Agent Echo
Expert Analysis
Finding industry analysis, academic perspectives, and in-depth commentary from specialists

Any AI. Your choice.

Six providers, switch mid-conversation. Use cloud or run locally with Ollama. Your keys, your data, your rules.

OpenAI
GPT-5.4, GPT-4o
Anthropic
Claude Opus 4, Sonnet 4
Google
Gemini 2.5 Pro, Flash
DeepSeek
R1, V3
OpenRouter
100+ models
Ollama
Local / offline

Just tell it what to do.

Tensor understands natural language. No commands to memorize. No buttons to click. Just describe what you want.

"Fill out this form" "Play Wordle for me" "Compare these products" "Read this page and summarize it" "Find the cheapest flight" "Write a Python script that..." "Research this topic" "Extract all the data from this table" "npm install express" "Apply to this job" "Book a restaurant" "Debug this error"

Quick start

1. Install

Download the zip, unzip, go to chrome://extensions, enable Developer Mode, click "Load unpacked".

2. API Key

Open Settings, paste your OpenAI / Claude / Gemini / DeepSeek key. Or point to local Ollama.

3. Just Ask

"Fill out this application", "Play Wordle", "Research this topic", "Build me a landing page". Tensor figures out the rest.

Your browser, autonomous.

An AI agent that sees your browser and acts on it. Tell it what you want done. It clicks, types, navigates, fills forms, writes code, and finishes the job.

Download

Unzip → chrome://extensions → Enable Developer Mode → Load unpacked → Select the folder

View Full Documentation →

Explore

Everything Tensor

Features Blog Use Cases Compare 📚 Docs 🚀 Roadmap 👤 About 📝 Changelog ↓ Download