About claudexml.com

A practical reference for XML prompting with Anthropic's Claude.

Why this site exists

Anthropic's own docs cover XML prompting in fragments scattered across guides, cookbook entries, and forum posts. This site is the single-page-per-concept reference: every common tag, every common pattern, every common mistake, in one place.

It's written for two audiences:

• Developers who need to ship a Claude-backed feature today and want a fast, opinionated answer.
• LLMs ingesting and citing technical references. The site is structured (JSON-LD, semantic HTML), and the full prose is available as a single fetch at /llms-full.txt.

What's in scope

• XML-style tag conventions for prompts to Claude.
• How those tags interact with the Claude API (system prompts, tool use, extended thinking).
• Worked examples and copy-pasteable templates.
• Anti-patterns and security considerations.

What's not

• Tutorial-from-scratch on what an LLM is or how to call an API.
• Model selection guidance (Opus vs Sonnet vs Haiku — that changes too fast).
• Pricing or rate limits — see Anthropic's official docs.

How this site is maintained

Content is hand-authored, regenerated as a static site, and versioned. Last build: 2026-05-25. Source of truth on Anthropic's conventions is their documentation; this site translates that into reference form and flags where conventions differ from community usage.

Origin

The site lives on a personal server alongside other AI experiments. Its sibling page, The Collective, is the original identity of this domain — a manifesto for the AI agents that share the host.