Pergel: Privacy-First Browser Engine with AI, Built for Agents

We built Pergel — an open-source browser engine that's privacy-first and AI-native. After months of development, we're excited to share it with the community.

What it does

• SSRF-guarded web extraction (browse, extract, crawl) via clean REST API
• Built-in AI mode: query fan-out, grounded citations, multi-model router
• Adblocking: EasyList + EasyPrivacy compiled to binary automata (ABP + uBlock syntax)
• Tracker blocking: 60+ tracker domains blocked natively
• Crash recovery: sessions auto-persist, never lose tabs
• Memory saver: inactive tabs hibernate automatically
• Zero telemetry, zero data collection, BYOK (your keys stay local)
• Web UI control panel + MCP tool server (WebMCP-ready)
• 8 languages, RTL support, session record/replay

What we DON'T do (on purpose)

• No stealth/anti-detection for bypassing paywalls or bot walls
• No residential proxies or CAPTCHA solving
• No user tracking, ever

Tech stack

Python + CDP (Chrome DevTools Protocol), PyQt6 WebEngine for the UI layer, FastAPI for the REST API. 802 tests, TDD-developed, MIT licensed.

Architecture

Three layers:

1. CDP Driver — Chromium control via Chrome DevTools Protocol
2. Brand Engine — PyQt6 WebEngine for the native UI
3. API Layer — FastAPI REST API for programmatic access

The adblocking engine uses Aho-Corasick + Wu-Manber algorithms on compiled binary automata for high-performance pattern matching.

Try it

git clone https://github.com/sametatakan0000-netizen/pergel
cd pergel
pip install -e .
python -m src.main

Would love feedback on the architecture — especially the SSRF-guarded extraction pipeline and the Pareto-dominance model router.

opensource #python #privacy #ai #browser