getmaxun/maxun

Maxun – Turn Any Website Into a Structured API

What it is

• An open‑source, no‑code platform that lets you collect data from the web and turn it into clean, structured outputs (JSON, CSV, Markdown, etc.).
• It bundles a visual UI, a set of “robots” (pre‑built automation agents), a CLI, and an SDK so you can run the same workflows locally, in Docker, or on the hosted service.

Why it matters for AI / ML

• Modern LLM applications need high‑quality, well‑structured data. Maxun can scrape sites, crawl whole domains, or extract information from PDFs and images, then output the data in a format ready for prompting or fine‑tuning.
• It offers an LLM‑powered extraction mode: you describe the fields you want in natural language and the system uses a language model to locate and return them, reducing the need for hand‑crafted selectors.

Key components

Component	What it does	AI‑related feature
Extract	Point‑and‑click or record a browsing session, then replay it to pull structured data.	AI Mode – use an LLM to specify what to extract instead of writing selectors.
Scrape	Saves a full page as clean Markdown/HTML and can capture screenshots.	Outputs LLM‑ready Markdown for downstream prompting.
Crawl	Traverses an entire site, respecting scope rules, and extracts data from each page.	Can feed large corpora into vector stores or training pipelines.
Search	Automates web searches and collects the results, with time‑based filters.	Useful for gathering fresh data for retrieval‑augmented generation.
Document Extraction	Parses PDFs, DOCX, XLSX, CSV and runs OCR on images.	Turns scanned documents into structured text for LLM ingestion.
SDK / CLI	Programmatic control, scheduling, and robot management.	Enables integration into custom AI pipelines or CI/CD.

How you use it

1. Hosted quick‑start – go to https://app.maxun.dev and create a robot via the web UI.
2. Self‑host – run the service locally with Docker Compose or a plain install (see the docs). This gives you full control over data and compute.
3. Automation – schedule robots to run periodically, expose the extracted data via a REST endpoint, or push directly to Google Sheets/Airtable.
4. LLM integration – the platform can output data in the Model Context Protocol (MCP) format, making it straightforward to feed results into LLM prompts or retrieval‑augmented pipelines.

Who it’s for

• Data engineers & ML teams that need a reliable way to harvest web data without writing scrapers from scratch.
• Product builders creating AI agents that require up‑to‑date web information.
• Researchers gathering large corpora for training or evaluation.
• Non‑technical users who can use the visual recorder to build extraction flows.

License

• AGPL‑v3, meaning the code is free to use and modify, but any network‑deployed service must share its source changes.

Bottom line Maxun bridges the gap between raw web content and the clean, structured datasets that modern AI systems need, offering both a no‑code UI and developer‑friendly tools for automation and integration.