# llms.txt generator

> The machine-readable front door to any website — generated on demand. Point the tool at a URL and it reads the sitemap (or crawls the homepage), lifts each page's title and summary, and assembles a spec-compliant llms.txt. An offering from crossinginto.ai. This file is the tool describing itself, in the format it produces.

## Use it

- [Generator](https://llmstxt.crossinginto.ai/): Paste a URL, watch it build in the open, copy or download the result.
- [API — generate (streaming)](https://llmstxt.crossinginto.ai/api/generate): POST JSON `{"url","options":{"maxPages","force"}}`; responds with a streamed NDJSON provenance log ending in a `done` event carrying the llms.txt and stats.
- [API — raw (curl)](https://llmstxt.crossinginto.ai/api/llms.txt): GET `?url=https://example.com` returns the llms.txt as `text/plain`, cache-first.

## How it works

- [Discover](https://llmstxt.org): Read sitemap.xml when present; otherwise follow same-origin links from the homepage, within a page budget.
- [Read](https://llmstxt.crossinginto.ai/): Fetch each page and extract only its title, description, and first heading — nothing else leaves the source site.
- [Assemble](https://llmstxt.org): Group pages by path into sections and write the file to the llms.txt specification.

## About

- [crossinginto.ai](https://crossinginto.ai): Crossing into machine intelligence — governance, provenance, memory. This tool is an offering under that brand.
- [The llms.txt proposal](https://llmstxt.org): The convention this tool implements.
- [Built by Matthew Williamson at Clevyr](https://crossinginto.ai): Design and engineering.

<!-- generated by hand, in the format llmstxt.crossinginto.ai produces · hello, machine -->