searchts Challenges AI’s Web Blind Spot

searchts targets a real reliability gap: agents often fail not because information is unavailable, but because their fetch method gets blocked. The project is compelling as personal-scale infrastructure, though its claims should be read alongside its residential-IP assumptions and CAPTCHA limitations.

• –Escalating from `curl_cffi` to Jina Reader and stealth Chromium keeps common reads fast and lightweight.
• –MCP, CLI, Python-library, and Claude Code integrations make it practical across agent workflows.
• –Support for AI-chat share links is a smart specialization that generic scrapers frequently miss.
• –Its benchmark reports 12/12 reads, but results depend on a residential connection and do not establish broad success against the hardest bot walls. [Benchmark](https://github.com/capad-xyz/searchts/blob/main/docs/scorecard.md)
• –The human-in-the-loop CAPTCHA fallback is an honest boundary, while default Jina routing introduces a third-party privacy tradeoff. [Project overview](https://capad.fyi/work/searchts)