browser-search  by Johell1NS

AI agent skill for robust web search and browsing

Created 2 weeks ago

New!

328 stars

Top 82.9% on SourcePulse

GitHubView on GitHub
Project Summary

Summary

This project offers a self-hosted, free, and unlimited "skill" for AI agents to perform web searches and browsing. It tackles anti-bot systems by orchestrating SearXNG (search), Camofox (standard browse), and CloakBrowser (stealth browse). The key benefit is an anti-hallucination system ensuring reliable, cost-free access to live web data for AI agents.

How It Works

The system first uses a Dockerized SearXNG for metasearch, returning URLs. AI agents then select Camofox (Dockerized Firefox via REST API) for standard browsing, leveraging Readability.js for efficient data extraction. If anti-bot protections (Cloudflare, Akamai) are detected, it automatically escalates to CloakBrowser (npm Chromium-based stealth browser) which uses advanced fingerprinting and challenge detection to bypass these barriers.

Quick Start & Requirements

  • Install Skill: npx skills add Johell1NS/browser-search for AI agents.
  • Infrastructure: Clone repo, npm install, run Docker for SearXNG (localhost:8080) and Camofox (localhost:9377).
  • Prerequisites: Docker, Node.js/npm. Lightweight, tested on Raspberry Pi.
  • Docs: SearXNG (docs.searxng.org), Camofox (github.com/jo-inc/camofox-browser).

Highlighted Details

  • Anti-Hallucination: Enforces "search first, answer second" workflow for verified, live web data.
  • Unlimited & Free: Fully self-hosted, no API keys, subscriptions, or rate limits.
  • Automatic Escalation: Seamlessly switches between Camofox and CloakBrowser based on site protection.
  • Advanced Stealth: CloakBrowser bypasses major anti-bot systems (Cloudflare Turnstile, FingerprintJS) with human-level scores.

Maintenance & Community

Encourages community contributions via GitHub issues and PRs. Open-source nature implies active development, though specific community channels are not detailed.

Licensing & Compatibility

  • License: MIT.
  • Compatibility: Works with various AI agents. MIT license generally permits commercial use.

Limitations & Caveats

  • Unsupported: Does not access social media requiring logins, bypass paywalls, or download files.
  • Security: Residual risks include prompt injection, supply chain concerns (proprietary Chromium binary), and general browser automation risks. Isolated environments recommended.
Health Check
Last Commit

4 days ago

Responsiveness

Inactive

Pull Requests (30d)
3
Issues (30d)
3
Star History
329 stars in the last 19 days

Explore Similar Projects

Feedback? Help us improve.