100% Local & Private

Chat With Any Webpage
Using Your Own AI

No switching apps. Click and ask — right where you are. A Chrome side panel that connects any webpage to your local AI models. Select text or parse the full page. Your data never leaves your computer.

Install Extension Setup Guide
4 Providers
Side Panel Always On
100% Private
Free To Use
PageGrok chatting with a selected section of a webpage using a local AI side panel
Features

Read and Chat,
Without Leaving the Page

A side panel that stays open while you browse. Select text or parse the full page. Your models, your data, your rules.

Side Panel

Opens alongside the page, not as a fragile popup. Stays open while you scroll, read, or copy text. True split-screen reading & chatting.

Text Selection

Select specific text on the page and chat about just that part. No need to process the entire page when you only care about one section.

Smart Suggestions

After parsing, AI generates relevant follow-up questions based on the actual content. Just click to ask — no blank-page anxiety.

Multi-Provider

Ollama, LM Studio, oMLX, or any OpenAI-compatible API. Use whatever models you have installed — Llama 3, DeepSeek, Qwen, Mistral.

Privacy by Design

All processing happens on your machine. No cloud, no accounts, no tracking. Some content you just don't want cloud AI to see.

Markdown Rendering

AI replies with formatted Markdown: code blocks with syntax highlighting, tables, lists. Copy or re-edit any message.

How It Works

Three Steps. That's It.

No accounts, no setup wizards. Make sure a supported model backend is running, then install and go.

Requires a model backend: run Ollama, LM Studio, or oMLX locally — or plug in any OpenAI-compatible API.
1

Browse Any Page

Navigate to any webpage — documentation, articles, papers, or code repositories.

$ ollama serve
Listening on 127.0.0.1:11434
# or LM Studio / oMLX / external API
2

Click & Parse

Click the PageGrok icon — a side panel opens alongside the page. Parse the full page, or select specific text to focus on.

PageGrok > Side Panel
Extracted 12,847 chars
or: Select text to focus
3

Chat with AI

Ask questions about the content. Get Markdown-formatted answers from your local LLM, with smart follow-up suggestions.

You: Summarize this page
AI: This page covers...
[3 suggestions generated]
Privacy

Your Data Stays
on Your Machine

Some content is sensitive — internal docs, unpublished papers, competitor analysis. PageGrok processes everything locally. Your page content and conversations are never collected — only anonymous performance stats, which you can turn off.

No Cloud

Zero network requests to external servers. All API calls go to localhost only.

No API Keys

No accounts, no tokens, no subscription for local providers. Install a model, install the extension, done.

Self-Controlled

Not affected by network restrictions or service availability. Your local model runs wherever your computer runs.

PageGrok keeping webpage conversations inside a protected local AI environment
Who It's For

Built for Work That
Can't Go to the Cloud

If what you read is sensitive, PageGrok keeps both the AI and your data on your own machine.

Research, Medical & Legal

Highly sensitive material you can't paste into cloud AI. A local model handles summaries and Q&A — nothing leaves your device.

Enterprise & Compliance

Point it at an internal Ollama instance. Staff get AI help on confidential docs and competitor analysis with no data-exfiltration risk.

Local AI Tinkerers & Developers

No app switching. Chat about the page you're on using the models you already run — Llama, DeepSeek, Qwen, Mistral.

Providers

Your Models, Your Choice

Four built-in provider channels. Use whatever AI you have running locally, or connect to an external API.

OL

Ollama

Auto-discovers loaded models. The most popular local AI runtime.

LM

LM Studio

Connect to LM Studio's local server. Auto-discovers loaded models.

MX

oMLX

Apple Silicon native inference via oMLX. Auto-discovers loaded models.

AP

External API

Any OpenAI-compatible endpoint. Connect to cloud providers when needed.

Llama 3
DeepSeek
Qwen
Mistral
Gemma
Phi-3
CodeLlama
Vicuna

... and any other model available in your local provider's library.

Under the Hood

Built for Developers

Clean architecture. Minimal footprint. Maximum transparency.

architecture
// Side Panel + multi-provider. Zero dependencies. popup.js (Side Panel UI / chat / suggestions) |-- chrome.tabs.sendMessage | --> content.js (DOM extract + selection) | |-- chrome.runtime.sendMessage --> background.js (multi-provider proxy) |-- Ollama 127.0.0.1:11434 |-- LM Studio 127.0.0.1:1234 |-- oMLX 127.0.0.1:8080 |-- Ext API user-defined

Manifest V3

Built on the latest Chrome extension platform with declarativeNetRequest for secure CORS handling.

Side Panel API

Uses Chrome's Side Panel API for a persistent chat interface that stays open alongside the page.

Shadow DOM Traversal

Recursive TreeWalker penetrates Shadow DOM boundaries, ensuring content extraction works on modern SPAs.

Multilingual UI

4 languages built-in: English, Chinese, Japanese, Russian. One-click switching.

Changelog

What's New

Track the latest updates and improvements to PageGrok.

2026-07-12

v1.0.7 Whole-Session Saving & Fresh Look

  • Saving now keeps the whole conversation across page switches; re-saving updates the same record instead of duplicating.
  • Saved chats show their source page, and history view links back to the original article.
  • Two-level copy: copy a single reply, or export the full conversation as Markdown with source URL and time.
  • Reopen the side panel and restore your last unsaved session with one click.
  • Visual refresh: deep-green accent on a paper-tone light theme, unified across extension and site.
  • Smart suggestions follow the UI language; the settings icon is now a proper gear.
2026-07-11

v1.0.6 Streaming Chat & Anonymous Stats

  • Streaming responses for Ollama NDJSON and OpenAI-compatible SSE, with automatic fallback.
  • Anonymous performance telemetry can be disabled and never includes page content, URLs, or chat text.
  • Response stats now show first-token latency, token speed, and cached tokens.
  • Long-page context is capped and recent-turn trimming reduces repeated token cost.
  • Error messages are localized by code across Chinese, English, Japanese, and Russian.
  • Model-not-found state now opens Settings directly from the selector.
2026-05-25

v1.0.5 Smart Automation & Chat History

  • Auto-detect local LLM services (Ollama, LM Studio, oMLX) on startup with configurable toggle.
  • Model warm-up: automatically pre-load model on launch for faster first response.
  • Chat history: save, browse, export past conversations across sessions.
  • Settings toggles for auto-context loading, auto-detection, and auto warm-up.
  • Model-not-found messages now guide users to Settings (fully i18n across 4 languages).
  • Language switcher as dropdown, Dark/Light theme toggle, Japanese & Russian support.
  • Settings panel redesigned with sections, help tooltips, and experimental features area.
  • Response stats & token count moved to experimental toggle (off by default).
  • Model warm-up now shows real-time progress and timing feedback in chat.
2026-05-12

v1.0.4 Side Panel & Multi-Provider

  • Switched from popup to Side Panel for persistent split-screen reading & chatting.
  • Added LM Studio and oMLX local inference support with auto model discovery.
  • Added interactive text selection mode for targeted content extraction.
  • AI replies now rendered with Markdown formatting and syntax-highlighted code blocks.
  • Smart follow-up suggestions: AI generates 3 relevant questions after parsing.
  • Message actions: copy and re-edit any AI reply or user message.
2026-03-14

v1.0.3 Cloud API & Region Selection

  • Added external OpenAI-compatible API support for cloud LLM providers.
  • Added interactive region selection mode for targeted content extraction.
  • Addressed Chrome Web Store review feedback and updated permissions.
2026-03-12

v1.0.2 Privacy & Performance

  • Switched content script to on-demand injection for better performance.
  • Added comprehensive privacy policy page.
  • Resolved Chinese encoding issues.
2026-03-10

v1.0.5 Initial Release

  • Core feature: Chat with any webpage using local Ollama models.
  • Smart extraction with heuristic pruning and Shadow DOM traversal.
  • Bilingual UI (English and Chinese).

Ready to Chat With
Your Webpages?

Install PageGrok and start asking your local AI about any webpage. No switching apps, no data upload, no cost.

Get PageGrok Set Up Your Backend