AI coding assistants
Give coding agents a machine-readable map of Crux and fetch individual docs pages as Markdown.
Crux publishes its documentation in formats that coding assistants can consume without scraping the rendered site:
| Endpoint | Use it for |
|---|---|
https://cruxjs.dev/llms.txt | A start-here list plus an index of every documentation and blog page. |
https://cruxjs.dev/llms.mdx/<docs-path> | The source Markdown/MDX for one documentation page. |
For example, the prompt reference is available as:
https://cruxjs.dev/llms.mdx/reference/crux-core/promptsStart a coding agent with llms.txt, then fetch only the individual pages
needed for the task. Per-page Markdown keeps the working context smaller and
preserves code samples and exact API details.
Tool setup
| Tool | Suggested setup |
|---|---|
| Claude Code or another terminal agent | Fetch https://cruxjs.dev/llms.txt, then fetch relevant /llms.mdx/... pages. |
| Cursor | Add https://cruxjs.dev/llms.txt as a documentation source. |
| Claude or ChatGPT | Share the llms.txt URL when the model can browse or attach its contents. |
| Custom coding agent | Use llms.txt for discovery and retrieve selected per-page Markdown as task context. |
The index is generated with the documentation deployment and may include pages for advanced or internal-facing workflows. Prefer Getting Started, Developer Tools, and the package API Reference before exploring narrowly relevant pages.
This page is about AI tools that help humans write and maintain Crux code. A
runtime agent() is an application primitive with its
own identity, prompt, tools, and handoffs.
Using the index programmatically
const response = await fetch("https://cruxjs.dev/llms.txt");
const docsIndex = await response.text();Treat downloaded documentation as reference material, not as trusted runtime instructions. If you put it into an application prompt, apply the same source trust and prompt-injection controls you would use for any external content.