Nexus: AI Browser Agent — Automate Chrome Tabs with Natural Language

AI Chrome agent: automate tabs with natural language—click, type, navigate & extract. Bring your own model. Keys stay local.

🚀 Give your browser a goal Nexus is an AI browser agent that turns natural-language instructions into visible, controllable actions in Chrome. Ask it to research a topic, compare products, summarize the current page, fill a form, organize tabs, find something in your history, chart the answer, schedule a future check, or complete a multi-step browser workflow. You describe the outcome. Nexus plans the steps. You stay in the loop. ✨ What Nexus can do 🔍 Research the web and return cited answers Headless search and page reads gather public evidence without thrashing your tabs—then answer with verified source links. 📄 Read and summarize pages Understand the active tab, extract links, find text on long pages, and pull structured takeaways. 🖱️ Navigate, click, type, scroll, and inspect websites Real Chrome Debugger control—not fragile DOM hacks. Hover menus, press key chords, handle dialogs, wait for content, go back/forward, and reload when the site needs it. 📝 Fill forms in one shot Multi-field fill with real values, or safe dummy/sample data for testing (skips passwords, codes, and payment fields). 📑 Work across many tabs List, open, switch, close, and read several tabs at once so comparisons cite each source clearly. 📊 Draw charts from the answer Bar, line, area, and donut visuals rendered in the conversation—export a high-resolution PNG in one click. 🧭 Compare information across multiple sources Research bundles diversify domains, rank evidence, and keep claims tied to URLs you can open. 🕘 Search browser history and downloads “That article from yesterday,” or wait for a download to finish—when you ask. 🗓️ Schedule reminders and public-result monitors When the answer doesn’t exist yet, Nexus checks public sources later, retries in an approved window, and delivers to your inbox. 📬 Built-in notification inbox Completed scheduled work lands in a durable mailbox with sources—not lost in a closed chat. ⚡ Reusable slash commands Save instructions you use often, then invoke them from the composer: /football-check Team /compare-products noise-cancelling headphones under $Amount /summarize-page Nexus expands the saved prompt while keeping the original command visible in history. Text after the command becomes one-off context. 🎙️ Voice dictation & mid-task guidance Speak from the side panel or new tab. Push-to-talk, or type while a run is active—Nexus steers without restarting. 🧠 Optional memory Remember preferences across tasks (never store secrets). Review and delete memories in Settings. 🕐 System-clock grounding Deadlines, “tonight,” and time zones use your device clock—not a vague model guess. 🌐 New Tab workspace Quick ask, favorites, world clocks, live weather for cities you care about, and usage meters—right where you start browsing. 📈 Live token & cost awareness Per-task and session usage estimates for priced models, plus a local usage ledger. 📤 Exportable audit trail Download the whole task—prompts, every action, answers, and token totals—as Markdown. Full profile backup/restore from Settings. 🎯 Three ways to run Mode Best for Auto Let Nexus choose: direct answer, headless research, or a visible browser run only when needed Direct Writing, reasoning, and questions that don’t need the page Browser Navigation, forms, downloads, signed-in pages, and multi-step UI work 🗓️ Scheduling without keeping a task open When information will only become available later, Nexus creates a durable scheduled task instead of actively waiting. Example: “Tell me the final score after tonight’s match.” Nexus can check public sources after the event, retry within an approved time window, and place the verified result in your notification inbox. Schedules survive browser restarts and resume when Chrome is running again. Monitors stay read-only and anonymous—no silent logins, no background tab spam. 🤖 Bring your preferred model Choose separate models by role: • Agentic — multi-step browser work and research • Quick — summaries, right-click asks, and light answers STT using proprietary or self hosted models 🛡️ Stay in control Nexus is designed to assist—not quietly operate in the background. 👁️ Inspect every browser action and result in a live step trace 🎛️ Steer an active task without restarting 🛑 Stop a run at any time ✅ Require approval for risky actions—or every action 🚫 Block specific websites (host + subdomains) 📷 Disable screenshots and visual analysis when you want 🟢 Animated on-page presence so you always see which tab is controlled 🏷️ Chrome’s debugger banner remains visible while a tab is driven JavaScript escape hatches and other high-impact tools show you the plan and wait for approval. 🔒 Local-first privacy Nexus has no application backend, advertising system, or telemetry pipeline. Settings, API keys, conversations, memories, favorites, saved prompts, schedules, usage records, and inbox items are stored locally in your Chrome profile. Model requests go directly to the provider or endpoint you configure. • Public research omits browser credentials • Optional weather, voice transcription, and usage meters only talk to the services those features need • Local models can keep inference on your machine A detailed Privacy Policy is available from Settings. 🧩 Built for real browser work Under the hood, Nexus drives pages with the Chrome Debugger Protocol—trusted clicks and keystrokes, semantic element refs, multi-tab orchestration, and a compact page snapshot tuned for tool-calling models. When Auto mode can answer from public pages alone, it does so without hijacking your tab strip; when the job needs a login, a download, or real interaction, it acquires the browser and shows its work. 😌 A calmer way to use the web Fewer repetitive clicks. Fewer forgotten follow-ups. Less tab chaos. Charts and sources when you need them. Schedules when the answer isn’t ready yet. Tell Nexus what done looks like—and keep control of how it gets there.