A no-code scraper that records your clicks and drafts LLMs to finish the job

What it does

Maxun is a self-hostable web data platform that automates extraction, scraping, crawling, and search through browser emulation. You can record interactions like a macro, describe targets in natural language for LLM-powered extraction, or convert entire pages and sites into Markdown, HTML, or spreadsheets. It handles pagination, scrolling, and authentication, then exposes the results via REST endpoints or direct exports to Google Sheets and Airtable.

The interesting bit

The project straddles two eras of automation: it has a point-and-click recorder for deterministic workflows, but also leans on LLMs to parse and extract data when layouts are unpredictable. It even speaks MCP (Model Context Protocol), positioning itself as infrastructure for AI agents rather than just a standalone tool.

Key highlights

• Dual extraction modes: deterministic recorder macros and natural-language LLM extraction
• Built-in crawling, automated search with time filters, and full-page scraping to clean Markdown
• REST API and spreadsheet exports; also an SDK and CLI for programmatic control
• Handles logins, pagination, and claims auto-recovery when site layouts change
• Self-hostable via Docker or local setup; licensed under AGPLv3

Caveats

• The README explicitly states the project is in “early stages of development,” so expect rough edges and shifting APIs
• AGPLv3 licensing means commercial users need to be comfortable with copyleft obligations

Verdict

Good fit for data engineers and founders who need quick, repeatable web data without maintaining a fleet of custom scrapers. Skip it if you need a battle-tested, enterprise-grade compliance layer or if copyleft licensing is a dealbreaker.