XiaoyaoClaw Web Clipper
A "knowledge feeder" for the web — send a link, get a clean Markdown with frontmatter in your local knowledge/. Dual-engine extraction, CN-friendly, batch dedup, fully local.
Product Introduction
OpenClaw Web Clipper fixes four web-clipping pains: bookmark hoarding you never reopen, copy-paste into notes with broken format and missing source/date, cloud clippers (Evernote/Cubox) that hold your data and charge a fee, and saved pages that can’t feed an AI KB in unified format. Solution: one command, web → clean local Markdown. readability-lxml (fast) + trafilatura (academic-grade) dual-engine fallback chain; bs4 as last resort. WeChat-Zhihu-CSDN container aware; GBK/UTF-8 auto-detect; CN-safe filenames. URL-list batch clipping with summary report; .clips-index.json auto-skips duplicates. Full frontmatter (title/source/author/date/tags), Obsidian-compatible. Defaults to knowledge/clippings/ — pair with KB Retriever for instant searchable index. Windows / macOS parity. No external API, no data leaves the machine.
Core Features
- Dual-engine extraction — readability-lxml (fast) + trafilatura (academic), auto fallback on low quality, bs4 last resort
- CN-friendly — WeChat/Zhihu/CSDN container aware; GBK/UTF-8 auto-detect; safe CN filenames
- Batch clipping — one URL list, full batch + summary report (ok / skipped / failed)
- Auto dedup — .clips-index.json skips duplicates automatically
- Full frontmatter — title / source / author / date / tags, Obsidian-compatible
- KB loop — defaults to knowledge/clippings/, pair with kb-retriever for instant search
- Cross-platform — Windows / macOS parity (pure Python)
- Fully local — no external API, data never leaves your machine