100% On-Device AI • Chrome Native Side Panel

Long articles in.
Structured takeaways out.
Zero cloud servers.

Brevira extracts article text and generates bullet-point summaries directly on your computer's local hardware with Chrome's built-in Gemini Nano model. No subscriptions, no accounts, and not a single byte leaves your browser.

Available for desktop Google Chrome on Windows, macOS, Linux, and Chromebook Plus.

  • 0 Cloud API Calls
  • 0 Tracking Cookies
  • 100% Free Utility
smithsonianmag.com | Brevira Side Panel
Authentic Google Chrome capture showing Brevira active in the native Side Panel on Smithsonian Magazine with structured key takeaway bullets
FOUNDER'S DISPATCH

Why I built Brevira

Every week, long-form journalism, research essays, and financial reports get buried in browser tabs because nobody has 35 minutes between meetings to sift through 8,000 words.

The existing reading assistants all made the same unacceptable trade-off: they send your reading history, private articles, and internal intranet pages to remote cloud servers—and then ask you for a $20/month subscription to cover their GPU bills.

When Google introduced built-in Gemini Nano inside desktop Chrome, the math changed. Your computer already has the compute power to summarize articles locally. There is no cloud server to maintain, no third-party API bill to subsidize, and no reason to store your reading habits.

Brevira is built as a focused, local utility. It does one job with high precision: strips clutter with Mozilla Readability, runs local on-device synthesis, and hands you structured Markdown takeaways directly in your side panel.

Pranav Chikanis Creator of Brevira • Built with Google Chrome Built-in AI Questions? Write to me at brevira.support@gmail.com
SIMPLE 3-STEP FLOW

How It Works

No copying and pasting into chat windows. No tab juggling. Just your article and your notes side-by-side.

STEP 01

Open any long-form article in Chrome

Navigate to an investigative report, academic preprint, or technical essay. You never have to highlight text, copy-paste snippets into separate tabs, or fiddle with external document converters.

Clean compatibility: Works on standard web articles containing readable narrative text (at least 200 characters).
smithsonianmag.com/history/the-untold-story...
Chrome window displaying the source article undergoing non-destructive local content analysis
chrome://side-panel | Brevira Docked
Brevira native side panel docked cleanly next to active article
STEP 02

Open Brevira in Chrome's native Side Panel

Click the Brevira icon in your browser toolbar. Unlike modal overlays that vanish the second you click back into the article, Brevira lives in Chrome's native Side Panel.

Pinned reading context: The side panel stays pinned beside your tab as you scroll, allowing you to cross-reference key claims with the source text instantly.
STEP 03

Mozilla Readability extracts; Gemini Nano streams

Brevira clones the webpage in memory (doc.cloneNode(true)) and runs Mozilla Readability to strip ads, sidebars, and trackers without touching your live DOM.

Chrome's local Gemini Nano model immediately streams bulleted takeaways into your side panel in real time.

Zero server roundtrips: Local hardware inference finishes in seconds with zero network latency.
Live Streaming | Gemini Nano Token Output
Live token streaming visualization in Brevira's Chrome side panel
UNDER THE HOOD

Technical Architecture & Privacy Boundary

Built on strict least-privilege standards. How Brevira extracts, summarizes, and formats text without cloud servers.

0 BYTES EXPORTED

Strict On-Device Privacy Boundary

Article narrative text is processed directly by Chrome's native Summarizer API on your local CPU/GPU. No text, URLs, or summaries are transmitted to external AI servers.

Active DOM Source Article
Readability In-Memory DOM Clone
Gemini Nano Local RAM Only
Side Panel Markdown Output
Zero network transmission to cloud AI vendors
MOZILLA READABILITY

Non-Destructive DOM Parsing

We clone the webpage in an isolated memory buffer before extracting narrative text. Your active webpage, forms, and video players are never modified.

// In-memory clone keeps active tab untouched
const clone = document.cloneNode(true);
const article = new Readability(clone).parse();
const cleanText = article.textContent;
CLEAN MARKDOWN

Clean Markdown & 1-Click Export

Summaries are formatted in GitHub Flavored Markdown with bold claims and bulleted insights. Ready for 1-click copy into Obsidian, Notion, Apple Notes, or Logseq.

### Key Takeaways
- **Core Thesis:** Quantized local models provide instant synthesis without server latency.
- **Data Boundary:** Zero bytes leave browser memory.
STRUCTURAL INTELLIGENCE

Adaptive Depth & Section Sampling

Long research essays can exceed single prompt windows. Rather than truncating solely from the end, Brevira samples proportionately across substantive headings so you don't miss conclusions.

Short (< 1,500 words) Medium (1,500 – 4,000 words) Long (4,000+ words)
COMMON QUESTIONS

Direct Answers to Real Questions

Honest explanations regarding local hardware execution, compatibility prerequisites, and permissions.

How can Brevira be completely free with no monthly subscription?

Brevira runs on Google Chrome's built-in AI model (Gemini Nano), which executes directly on your computer's local hardware. Because summarization runs on your machine with $0 marginal server compute cost, Brevira is offered completely free of charge without paywalls, subscriptions, or credit card forms.

Does Brevira transmit my reading history or article content to any servers?

No. Brevira operates under an on-device architecture. When you summarize a supported article, the text is extracted and processed locally inside your browser memory. Brevira does not transmit your article text, URLs, titles, or summaries to external AI servers or analytics platforms.

Will Brevira work on my device? What are the system requirements?

Brevira relies on Chrome's built-in AI (Gemini Nano) via Chrome's native Summarizer API. While modern processors, GPUs, or NPUs assist with machine learning workloads, Google Chrome directly manages model availability in your specific browser environment.

The most reliable way to find out is to install Brevira and check the setup card in the side panel. If Chrome needs to download model weights, Brevira guides you through the process.

What happens during the first-time setup?

If Chrome has not previously downloaded its built-in AI model components, Brevira will display a "Setup Required" card. Clicking "Download AI Model" prompts Chrome to download the necessary weights in the background. Once downloaded, Brevira is ready for local offline summarization.

Can Brevira summarize paywalled articles, local PDFs, or web apps?

Brevira extracts text from supported web articles rendered in your active browser tab using Mozilla Readability. If you have legitimate access to view the article text on your screen (such as through an active personal subscription), Brevira can summarize that visible text locally. Brevira does not bypass paywalls. Standalone local PDF files and interactive dashboards lacking narrative text are not currently supported.

Why does Brevira request activeTab, scripting, and sidePanel permissions?

Brevira strictly follows the principle of least privilege and avoids broad host permissions (<all_urls>). The activeTab permission grants temporary access to read the active webpage only when you explicitly click the icon. The scripting permission injects Mozilla Readability to isolate narrative text. The sidePanel permission allows Brevira to display its interface cleanly alongside your tab.

How do I contact support or join the community?

You can reach out directly by emailing brevira.support@gmail.com or by joining our official community Discord at https://discord.gg/86yTYQw3X3.

Start reading deeper in less time.

No accounts. No subscriptions. 100% on-device AI in Chrome's native Side Panel.

Available for desktop Google Chrome on Windows, macOS, Linux, and Chromebook Plus devices.