What do you use to turn websites into clean LLM context?

I’m launching webclaw here because I kept hitting the same problem while building agents and RAG workflows: getting a page is easy, but getting clean context from that page is not.

Raw HTML usually brings too much noise: nav, footers, cookie banners, duplicated layout text, scripts, missing JS-rendered content, and inconsistent structure.

webclaw is my attempt at solving that layer: scrape/crawl/map websites and return clean markdown, JSON, structured extraction, summaries, diffs, and MCP/CLI-friendly output.

I’m curious how people here handle this today.

If you’re building agents, RAG pipelines, or internal tools:

- Do you prefer markdown, JSON, or schema-based extraction?

- Are you using Firecrawl, Apify, Jina Reader, Crawl4AI, Browserless, custom Playwright, or something else?

- What usually breaks first: rendering, bot protection, noisy markdown, cost, or crawl quality?

Would love to compare notes and learn what your actual workflow looks like.