browser-search

browser-search

A skill for AI agents. OpenCode, Claude Code, Cursor, OpenClaw and beyond. Search the web with SearXNG, browse with Camofox, bypass protections with CloakBrowser. Anti-hallucination by design. All self-hosted, free, unlimited.

English · 简体中文 · Español · 日本語 · 한국어 · Português (Brasil) · Français · Deutsch · Русский · العربية · हिन्दी · Italiano

Why it exists

Today an AI agent browsing the web is like a thief in a balaclava sneaking around a police academy. The site protections block it, challenge it, reject it.

👮 browser-search flips the situation: your agent stops being the thief and becomes the chief of police. No more clumsy access attempts. It walks through every door because it has the right tools. SearXNG to search, Camofox to browse, CloakBrowser when the site plays hardball.

The skill enforces the exclusive use of deterministic scripts. This eliminates model hallucinations, even with the cheapest models. The 3 tools are described in natural language, but execution is rigid: the model can neither get the command wrong nor misinterpret the output. The result is guaranteed success on every query — the skill and its deterministic scripts guide the model to scour the web until it finds the answer.

1. SearXNG — metasearch engine for the search phase (multi-source, JSON)
2. Camofox — browser navigable via REST API for standard sites
3. CloakBrowser — stealth browser for anti-bot protected sites

The typical flow: the agent first searches with SearXNG, then browses the results with Camofox (or CloakBrowser if the site is protected).

Benefits

• 100% free, self-hosted, unlimited. No API keys to buy, no subscriptions, no rate limits. Everything runs on your machine, Docker and npm. Unlimited usage, zero cost.

• Lightweight, runs anywhere. Built and tested on a Raspberry Pi — if it runs there, it runs everywhere. Minimal resource consumption, no heavy infrastructure needed, runs 24/7 on low-power hardware.

• Search + browse in one kit. No manual integration needed. Searching and browsing are two distinct phases, both covered.

• Automatic navigation escalation. If Camofox gets blocked by Cloudflare/Akamai, the agent automatically switches to CloakBrowser.

• Smart performance. SearXNG for the search phase (milliseconds). Camofox and CloakBrowser are only used to browse the sites that actually need it.

• Automatic agent choice. The AI agent decides which tool to use: SearXNG for initial search, Camofox for browsing, CloakBrowser if the site is protected. Zero human intervention.

• Anti-hallucination by design. The skill’s Deep Research mode enforces a “search first, answer second” workflow: the agent must verify every factual claim against live web sources, cross-reference multiple angles, and never guess. No more made-up answers.

• Fully customizable. The SKILL.md is plain text. You can edit the core rules, add your own, remove what you don’t need. Adapt it to your workflow, your team, your standards.

• Native stealth. CloakBrowser automatically detects Cloudflare, Akamai, DataDome, Imperva, PerimeterX, and DDoS-Guard challenges, and waits for them to resolve before extracting content.

• Works with any agent. The SKILL.md is written for OpenCode, but the logic is identical for any AI agent. Same README, same package.json, everything works everywhere. Just ask your agent how to convert the skill for its environment.

Why it’s different

Most web tools for AI agents stop where the wall starts: when a site throws a Cloudflare or Akamai challenge, they hand back an error and the agent gives up. browser-search is built to walk through instead.

• Search → browse → bypass in one skill. No manual integration, no per-query costs, no API keys to buy.
• Three-tier escalation. SearXNG for milliseconds-fast search, Camofox for the ~90% of standard sites, CloakBrowser when the site plays hardball — automatically, with zero human intervention.
• Proven anti-bot results. CloakBrowser applies 58 C++ source-level patches and scores 0.9 on reCAPTCHA v3 (human-level, server-verified), clearing Cloudflare, Turnstile, DataDome, Akamai, Imperva, PerimeterX and DDoS-Guard — where other tools just return blocked_by_challenge.
• Free, private, self-hosted. SearXNG, Camofox and CloakBrowser all run on your machine. Zero cost, unlimited, nothing metered.

Where other tools give up, browser-search keeps going.

🏆 State of the art

These three tools were chosen because they represent the current state of the art available today. A skill like this is designed to evolve: when better tools emerge, updating the SKILL.md is all it takes to swap them in. 🔄

⭐ Star the repo and follow to stay up to date on new tools, flow improvements, and orchestration updates over time. 🚀

Architecture

┌─────────────────────────────────────────────────────────┐
│                    browser-search                       │
│                                                         │
│  ┌──────────────┐                                       │
│  │    Search    │                                       │
│  │              │                                       │
│  │  SearXNG     │   search engines → URLs               │
│  │  (Docker)    │   JSON results, fast                  │
│  │  :8080       │                                       │
│  └──────────────┘                                       │
│         │                                               │
│         │ results ready → to browse                     │
│         ↓                                               │
│  ┌─────────────────────────────────────┐                │
│  │           Browsing                  │                │
│  │                                     │                │
│  │  ┌──────────────┐                   │                │
│  │  │   Camofox    │  browser + REST   │                │
│  │  │  (Docker)    │  JS, click, eval  │                │
│  │  │  :9377       │                   │                │
│  │  └──────┬───────┘                   │                │
│  │         │                           │                │
│  │         │ if blocked                │                │
│  │         ↓                           │                │
│  │  ┌──────────────┐                   │                │
│  │  │ CloakBrowser │  stealth Chromium │                │
│  │  │   (npm)      │  anti-bot, proxy  │                │
│  │  └──────────────┘                   │                │
│  └─────────────────────────────────────┘                │
└─────────────────────────────────────────────────────────┘

How it works

Phase 1 — Search with SearXNG

Docker container on localhost:8080. Metasearch engine that queries Google, Wikipedia, Bing, DuckDuckGo and many others simultaneously. JSON output with titles, snippets, and URLs.

Example:

node scripts/searxng/searxng.mjs search "largest llm benchmark 2026"

The agent now has a list of URLs to visit and autonomously decides whether to browse them with Camofox or CloakBrowser based on the site.

Phase 2 — Browse with Camofox

Docker container on localhost:9377. Exposes a full Firefox browser through a REST API. The agent can create tabs, navigate, click, scroll, execute arbitrary JavaScript, and structure data.

Includes: Mozilla’s Readability.js for extracting clean articles, removing nav, sidebar, and ads (~70% token savings).

Main commands:

# Single-URL extraction (Readability.js, auto-fallback to snapshot)
node scripts/camofox/camofox.mjs readability "https://example.com"

# JavaScript evaluation
node scripts/camofox/camofox.mjs evaluate "https://example.com" "document.title"

# Accessibility snapshot
node scripts/camofox/camofox.mjs snapshot "https://example.com"

Phase 3 — Browse with CloakBrowser (when Camofox isn’t enough)

npm package based on Playwright + cloakbrowser. Launches a Chromium browser with advanced fingerprinting to bypass Cloudflare, Akamai, DataDome and other anti-bot systems. Automatic challenge detection with wait and retry.

Available scripts:

• cloak-fetch.mjs — universal fetch with challenge detection
• cloak-script.mjs — custom Playwright script execution

Example:

node scripts/cloak/cloak-fetch.mjs "https://protected-site.com"
node scripts/cloak/cloak-fetch.mjs "https://protected-site.com" --proxy socks5://... --geoip

# Markdown output (requires: pip install markitdown)
node scripts/cloak/cloak-fetch.mjs "https://example.com" --format markdown

Version note: browser-search pins CloakBrowser ^0.6.3 and Playwright Core 1.64.1. CloakBrowser 0.6 requires Node.js 20.19+ or 22.12+. The skill uses the free CloakBrowser binary (Chromium 146; 145 on macOS), which needs no license key; the wrapper updates (Linux arm64 support, safer binary download, stable fingerprint per persistent profile, more reliable humanized actions) apply to it too. A free key (GitHub sign-in via cloakbrowser login) unlocks the newest binary with one concurrent session; the skill neither uses nor requires it.

Why both Camofox and CloakBrowser?

Because speed and stealth are a tradeoff, and the right tool depends on the site.

Camofox — fast, structured, persistent. Camofox wraps Camoufox (a C+±level Firefox fork) in a REST API with an always-warm browser. After a ~1-3s cold start, every request is near-instant. Its accessibility snapshots are ~90% smaller than raw HTML, with stable element refs (e1, e2, …) for reliable interaction. It handles the ~90% of sites that don’t use advanced anti-bot protection: articles, docs, search engines, standard web pages.

CloakBrowser — stealth, anti-bot, on-demand. CloakBrowser launches a fresh Chromium instance per request (~1-3s startup each time). It uses advanced fingerprinting, proxy support, geoip, and automatic challenge detection to bypass Cloudflare, Akamai, DataDome, Imperva, PerimeterX, and DDoS-Guard. It is the last resort for the ~10% of sites that block Camofox.

Real-world numbers:

• CloakBrowser applies 58 C++ source-level patches and scores 0.9 reCAPTCHA v3 (human-level, server-verified), passing all major anti-bot tests including Cloudflare Turnstile and FingerprintJS [²]
• Camofox cold start: ~1-3s (one-time, then ~0ms per request via warm REST API) [³]
• Playwright/Chromium cold start: ~0.5-6s (every launch, varies by environment) [⁴]

Camofox handles the fast path. CloakBrowser handles the edge cases. Together they cover the entire web with no gaps. The agent decides which to use.

Sources

¹ “Camoufox Vs Playwright Stealth: Complete Comparison & Alternatives (2026)” — blog.send.win ² CloakBrowser README — github.com/cloakhq/cloakbrowser ³ camoufox-pi README (cold start comparison) — github.com/MonsieurBarti/camoufox-pi ⁴ Playwright issue #4345 (launch time variability) — github.com/microsoft/playwright/issues/4345

Installation

Step 1 — Install the skill

Install the skill definition into your AI agent with a single command:

npx skills add Johell1NS/browser-search

This works with 70+ AI agents including OpenCode, Claude Code, Cursor, GitHub Copilot, and more.

Required after npx skills add: CloakBrowser is an npm dependency (cloakbrowser + playwright-core). The skills installer copies files but does not run npm install automatically. You must install it manually inside the skill directory:

# Find your skill directory (OpenCode example):
#   ~/.config/opencode/skills/browser-search
#   or .agents/skills/browser-search  (project-local)
cd ~/.config/opencode/skills/browser-search
npm install
node -e "import('cloakbrowser').then(c => c.ensureBinary())"
# Verify everything:
bash scripts/check.sh

Without this step, SearXNG and Camofox (Docker) will work, but smart-extract → CloakBrowser (--fallback / full-auto escalation) will fail with ERR_MODULE_NOT_FOUND: Cannot find package 'cloakbrowser' — run bash scripts/check.sh (see cloakbrowser not installed) or FAQ.md → Troubleshooting.

Step 2 — Set up the infrastructure (alternative: manual clone)

If you prefer to work from a git clone instead of npx skills add:

git clone https://github.com/Johell1NS/browser-search
cd browser-search
npm install
node -e "import('cloakbrowser').then(c => c.ensureBinary())"

CloakBrowser is installed by npm (requires npm install + binary download via ensureBinary). SearXNG and Camofox require separate Docker containers — make sure they are running before using the skill.

Show this README to your AI agent for a complete installation tailored to your environment and platform.

browser-search does not provide platform-specific docker-compose files or install scripts. Your AI agent reads the references below and adapts the setup to your OS, architecture, and environment automatically.

Services overview:

SearXNG configuration (required)

The default SearXNG install has JSON output disabled, and the skill needs it: without it the script fails with HTTP 403. In SearXNG’s settings.yml (the core-config folder in the official Docker setup) use:

use_default_settings: true
server:
  secret_key: "<random>"   # or set SEARXNG_SECRET
  limiter: false           # local instance: the limiter would block the script's JSON requests
search:
  formats: [html, json]

• use_default_settings: true keeps engines and defaults aligned with upstream on every update. To switch on extra engines, add entries such as engines: → - name: bing with disabled: false.
• Keep it local: bind SearXNG to 127.0.0.1 (SEARXNG_HOST=127.0.0.1 in .env).
• Valkey/Redis is not needed by the skill (SearXNG uses it only for the limiter, which stays off).
• If SearXNG is not at http://127.0.0.1:8080, set SEARXNG_URL.

For the AI agent — read these files:

Note: SKILL.md is written for OpenCode syntax (exec, node scripts). If your agent uses a different format (Claude Code, Cursor, etc.), read it and convert the commands to your agent’s syntax before using the skill.

Environment variables

What this skill does NOT do

• Social media. Instagram, Facebook, TikTok, LinkedIn, and Twitter/X require login. browser-search does not attempt to browse them.
• Download files. It is read-only (except for explicit screenshots).
• Bypass paywalls. Does not circumvent payment or login systems.

Security

browser-search includes multiple layers of security hardening:

Built-in protections

• SSRF prevention. URLs are validated before navigation — internal IPs (127.x, 10.x, 192.168.x, 169.254.x), cloud metadata endpoints, and .internal/.local TLDs are blocked. DNS resolution is also checked to prevent DNS rebinding attacks.
• Script sandbox. Custom scripts (cloak-script.mjs) run in a sandbox that restricts the Playwright API surface (only whitelisted methods on page, browser, context are accessible). Note: Node.js APIs remain available — for full isolation, a vm.Context would be required. Use --unsafe to bypass the sandbox and SSRF protection.
• Path traversal protection. --script paths must be within the skill directory. Absolute paths and ../ traversal are blocked.
• Rate limiting. 30 requests/minute by default to prevent accidental DoS or anti-bot triggers (use --no-rate-limit to disable).
• Secure filenames. Screenshots use random UUIDs instead of predictable timestamps.
• Stack trace suppression. Error output omits stack traces by default. Use --verbose for debugging.

Best practices

• API keys. Use environment variables ($CAMOFOX_API_KEY) or --env-file for Docker. Never paste keys on the command line — they appear in ps aux and shell history.
• Docker binding. Always use 127.0.0.1: prefix for port mapping (-p 127.0.0.1:9377:9377). Never expose to 0.0.0.0.
• URL encoding. Deterministic scripts handle encoding internally. No manual escaping needed.
• Version pinning. Dependencies use exact versions (no ^ caret ranges) and package-lock.json for reproducible builds.

Audit

Run bash scripts/audit.sh to verify the security posture of your installation.

Residual risks

• Prompt injection. An AI agent can be tricked into performing harmful actions. The tooling mitigates damage but cannot prevent a fully compromised agent.
• Supply chain. CloakBrowser downloads a Chromium binary from cloakbrowser.dev. The binary is SHA-256 verified but proprietary.
• Browser automation. Any tool with browser access has inherent risks. Run in isolated environments when possible.

Get involved

browser-search is open source and free. If you find it useful:

• ⭐ Star the repo — helps others discover it
• 🐛 Open an issue — report bugs or suggest features
• 🔀 Submit a PR — fix, improve, extend
• 💬 Share it — with your team, on Reddit, Twitter, Discord
• 🧠 Adapt it — fork it, tweak the SKILL.md, make it yours

Every contribution, no matter how small, makes this better.

FAQ

See FAQ.md for frequently asked questions about installation, architecture, design decisions, and common issues.

License

MIT