Privacy-first AI proofreader built for seamless in-browser editing

A small but persistent friction plagues many of us when writing online: the back-and-forth of copying text into a separate AI tool, waiting for a response, and pasting the result back into the original message. This tiny interruption breaks concentration, especially when it happens repeatedly across email drafts, social posts, or chat windows.

To solve this, one developer built PROOFER, a privacy-first Chrome extension that brings AI-powered proofreading directly into the web pages where you already type. Instead of leaving the context of your writing, you can select text, trigger the tool with a single click, and receive a revised version almost instantly—without exposing your content to external servers.

How PROOFER works: seamless editing without context switching

PROOFER integrates proofreading into the browser’s native workflow. When you’re composing a message or document, you simply highlight a sentence or paragraph and invoke the tool via a floating button, context menu, or keyboard shortcut. The extension sends only the selected text to your chosen AI model, streams the response back in real time, and displays a clean comparison between your original and the AI’s suggestion.

The core design principle is flexibility. PROOFER isn’t locked to a single AI provider. Users configure their own endpoint in the extension popup—supporting tools like LM Studio, Ollama, or any OpenAI-compatible API. This model-agnostic approach empowers users to run inference locally using Gemma 4, ensuring no sensitive text ever leaves their device.

Four editing modes are available to match different writing needs:

• Proofread: fixes typos, grammar, and punctuation
• Rephrase: improves clarity and flow
• Professional: polishes tone for formal or business contexts
• Shorten: condenses long text into a tighter version

Additional features enhance usability, including real-time streaming of responses, built-in connection testing, and domain-specific disabling via regex patterns to avoid interference on restricted sites.

Why privacy-first AI makes sense for everyday writing

Most AI writing tools require copying text into external tabs or cloud-based interfaces—platforms we don’t control. For sensitive drafts, work emails, or private conversations, this introduces unnecessary risk. PROOFER shifts the model inference to the user’s local environment, so the text never leaves the browser unless explicitly configured otherwise.

This local-first design aligns with the project’s goal: to make AI assistance feel immediate and trustworthy. By running Gemma 4 E2B locally through LM Studio or Ollama, users maintain full control over their data while still benefiting from fast, style-aware rewrites. The model responds quickly, follows instructions reliably, and avoids long delays that disrupt writing flow.

The developer behind PROOFER discovered its value firsthand while testing. Even casual conversations in Discord became smoother once the tool was available inline, proving that small friction points add up over time—and eliminating them can significantly improve daily productivity.

Setting up PROOFER: from local model to browser extension

Getting started with PROOFER is straightforward for users comfortable with local AI setups. After cloning the repository and installing dependencies via npm, users can load the extension as an unpacked Chrome or Edge extension. From there, configuration is done entirely through the extension popup:

npm install
npm run build

Once loaded, users configure their AI model endpoint by selecting a provider (LM Studio, Ollama, etc.), entering the base URL, adding an API key if required, and specifying the model name—such as Gemma 4 E2B.

After testing the connection, the extension is ready to use. On any webpage with an editable field, users can select text, trigger PROOFER, select an editing style, and review the streaming suggestions before accepting or discarding the changes. For maximum privacy, running the model locally ensures no third-party servers ever see the content.

The future of private, in-place AI editing

PROOFER demonstrates how focused AI tools can deliver real value when embedded directly into the user’s workflow. By prioritizing privacy, speed, and simplicity, it turns a minor annoyance into a seamless enhancement—one that stays out of the way until needed.

As local AI models become more accessible and efficient, extensions like this could redefine how we write online. No more context switching. No more waiting. Just faster, clearer, and more confident communication—all under the user’s control.