Auto-bypass for blocked websites in Claude Code — Phase 0→3 adaptive scheduler, no API keys
1.9k
Stars
245
Forks
2
Open issues
2
Contributors
AI Analysis
insane-search is a Claude Code plugin that bypasses access blocks to public websites by escalating through multiple retrieval strategies—public APIs, syndication feeds, TLS impersonation, and headless browsers—without requiring API keys or proxy setup. It is specialized for Claude Code users who need to extract public content from blocked or rate-limited platforms (X, Reddit, YouTube, LinkedIn, etc.); it is not a general-purpose web scraper and requires Claude Code as its runtime.
Inferred from signals mentioned in the README (tests, CI, type safety) — not a review of the actual code.
AI's overall editorial judgment — not an average of the bars above, can weigh other factors too.
Claude Code plugin that escalates through public routes to bypass access blocks—no API keys or setup
insane-search is a Claude Code plugin that automatically attempts multiple methods to access publicly available content when a direct fetch is blocked (403, CAPTCHA, WAF). It escalates through special public endpoints, syndication gateways, TLS impersonation, and headless browser techniques—all without requiring API keys or user credentials. Designed specifically for Claude Code users who need reliable access to public web content for research, summarization, and data gathering tasks. Adoption is growing rapidly among Claude Code users, but remains confined to that ecosystem.
Created in April 2026 as part of the gptaku-plugins ecosystem. The project addresses a specific friction point: Claude Code's default fetch functionality stops at access blocks, forcing users to manually work around them. insane-search automates that workaround by implementing an escalation pipeline.
Grew from 0 to 1,472 stars in ~2.5 months, with 176 stars in the last 7 days alone (as of 2026-06-28). This growth pattern suggests strong uptake among the Claude Code user base coinciding with increased adoption of Claude Code itself. The project is part of a larger ecosystem (gptaku-plugins), which likely amplifies discovery and adoption.
Adoption not verified in detail, but several indicators suggest real use: (1) 1,472 stars and 201 forks in a short timeframe is substantial for a Claude Code–specific plugin; (2) the plugin is listed in what appears to be an official marketplace mechanism (`/plugin marketplace add`); (3) recent spike of 176 stars in 7 days aligns with Claude Code rollouts or media attention; (4) users are encouraged to file issues and report platform-specific blocks, implying active bug reports and real-world testing. However, no public case studies, company endorsements, or quantified usage metrics are evident.
Based on README, the system uses a Phase 0→3 adaptive scheduler: Phase 0 attempts platform-specific public endpoints; Phase 1 tries lightweight probes (public APIs, syndication, URL variants); Phase 2 uses TLS impersonation with full browser identity (curl_cffi); Phase 3 deploys a real headless browser while monitoring internal JSON APIs. The design prioritizes fallback chains and detection of authentication boundaries. Likely implements dependency auto-install for tools like curl_cffi and yt-dlp.
not documented in README
Last push on 2026-06-28 (same as evaluation date) indicates active, ongoing development. The project is only ~2.5 months old but shows consistent momentum. No evidence of abandoned periods or stalled releases. README is multilingual (English, Korean, Chinese, Japanese, Spanish) suggesting active internationalization effort.
ADOPT IF: you use Claude Code and regularly hit access blocks on public content (Reddit, X, YouTube, etc.) and want automated workarounds without manual setup or API keys. AVOID IF: you need support for authenticated content, paywalled articles, or non-Claude Code environments; this is tightly coupled to Claude Code's plugin ecosystem. MONITOR IF: you're evaluating long-term stability—the project is young (2.5 months old) and growth may plateau once Claude Code adoption saturates, and maintenance dependency on curl_cffi and yt-dlp libraries means upstream breaks could affect it.
Independent dimensions
Mainstream potential
4/10
Technical importance
6/10
Adoption evidence
5/10
- Platform-specific brittle—sites change their public APIs and URL structures frequently; each update to a platform's structure may require code updates. README acknowledges this by encouraging issue reports.
- Tight coupling to Claude Code ecosystem—if Claude Code's plugin system changes or if Anthropic modifies fetch policies, the plugin may become obsolete or require significant refactoring.
- Legal and terms-of-service ambiguity—many platforms prohibit automated access even to public content. The README states it uses 'standard, documented techniques' but users could face account blocks or legal friction.
- Dependency on external tools—relies on curl_cffi, yt-dlp, and headless browser management; upstream breaks or API changes in these libraries could cascade into insane-search.
- Maturity unknowns—project is new (April 2026); no long-term stability track record. Error handling, edge case coverage, and real-world reliability at scale are unproven.
insane-search will likely remain a strong niche tool within the Claude Code ecosystem for 12–24 months. Growth will be constrained by Claude Code adoption ceilings; if Claude Code achieves broad enterprise adoption, insane-search becomes more valuable and may be ported to other AI coding platforms (Cursor, GitHub Copilot, etc.). Risk of fragmentation as platforms actively defend against automated access. Maintenance burden will likely increase as sites patch detection.
Newsletter
Get analyses like this every Monday
Free weekly digest of the most interesting open-source discoveries.
Languages
Information
- Language
- Python
- License
- MIT
- Last updated
- 1w ago
- Created
- 3mo ago
- Analyzed with
- anthropic/claude-haiku-4-5
Stars over time
Contributors over time
Top 100 contributors only — repos with more will plateau at 100.
Open issues
No open issues — clean slate.
Top contributors
Similar repos
eze-is/web-access
Web-Access is a skill for AI agents (Claude Code, Cursor, Gemini CLI) that adds...
Panniantong/Agent-Reach
Agent Reach is a Python CLI tool that provides AI agents with unified internet...
sarperavci/CloudflareBypassForScraping
This project provides a Python-based API service to bypass Cloudflare...
| Repository | Stars | Week Δ | Language | Score | Updated |
|---|---|---|---|---|---|
|
fivetaku/insane-search
this
|
1.9k | +221 | Python | 7/10 | 1w ago |
 decolua/9router
|
21.6k | — | JavaScript | 7/10 | 1h ago |
|
eze-is/web-access
|
8.2k | — | JavaScript | 7/10 | 2mo ago |
|
Panniantong/Agent-Reach
|
54.3k | — | Python | 8/10 | 1w ago |
|
sarperavci/CloudflareBypassForScraping
|
2.5k | — | Python | 7/10 | 1d ago |
 Alvin9999-newpac/fanqiang
|
2.4k | — | Shell | 3/10 | 6h ago |
Larger ecosystem (7,915 stars) and JavaScript-based; broader platform support but requires more manual setup. insane-search is Python-specific to Claude Code and prioritizes zero-config auto-escalation.
Similar star count (2,411) and Python language. Focuses narrowly on Cloudflare WAF; insane-search is broader (multiple platforms) and integrated into Claude Code rather than standalone.
Largest competitor (44,137 stars), Python, general-purpose agent toolkit. insane-search is narrower (Claude Code only) but more specialized for the specific problem of unblocking public web access.
JavaScript, routing-focused (18,648 stars). Different problem domain; insane-search is content-access specific.
insane-search is a plugin layer that extends, not replaces; the comparison shows clear gaps in the default behavior (403, CAPTCHA, media transcripts) that insane-search fills.
Phase 0→3 adaptive scheduler — details