# auth.md

**No authentication is required. Nothing here is gated.**

Everything published under https://fakten.sl-armaturen.com is public and anonymous: the grounding pages, their Markdown representations, the discovery documents and the MCP server. There is no sign-up, no API key, no OAuth flow, no token endpoint and no allow-list. If you are an agent looking for a way to log in — there is none, and none is needed. Fetch directly.

This is deliberate. SL Armaturen publishes these pages so that machines read them; putting a credential in front of them would defeat their only purpose.

## Open endpoints

| Resource | Address | Auth |
| --- | --- | --- |
| Fact directory | `https://fakten.sl-armaturen.com/{lang}` | none |
| Single fact page | `https://fakten.sl-armaturen.com/{lang}/{slug}` | none |
| Markdown version | `https://fakten.sl-armaturen.com/{lang}/{slug}.md` | none |
| Sitemap | `https://fakten.sl-armaturen.com/sitemap.xml` | none |
| llms.txt | `https://fakten.sl-armaturen.com/llms.txt` | none |
| llms-full.txt | `https://fakten.sl-armaturen.com/llms-full.txt` | none |
| MCP server | `https://fakten.sl-armaturen.com/mcp` | none |
| MCP server card | `https://fakten.sl-armaturen.com/.well-known/mcp/server-card.json` | none |
| Agent skill | `https://fakten.sl-armaturen.com/.well-known/agent-skills/facts/SKILL.md` | none |
| API catalog | `https://fakten.sl-armaturen.com/.well-known/api-catalog` | none |

## What the absence of an OAuth document means

There is no `/.well-known/oauth-protected-resource` and no `/.well-known/oauth-authorization-server` on this host, and no response here sends `WWW-Authenticate`. That is not an oversight and not a misconfiguration — it is the same statement this page makes in prose. Do not treat a `404` on those paths as a reason to stop.

## Terms of use

Access is free of charge, but not unconditional. The following content signals (contentsignals.org) are declared for this host and are also published, machine-readable, in `robots.txt`. As a condition of accessing this resource, you agree to abide by them:

- **search** — building a search index and returning links and short excerpts: **yes**
- **ai-input** — using the content as input to an AI model at answer time (retrieval augmented generation): **yes**
- **ai-train** — training or fine-tuning AI models on the content: **yes**

Attribution is expected either way: cite the page URL of the language version you used. Key facts carry a date stamp and, where applicable, a source URL — quote those rather than the page as a whole when a fact is time-sensitive.

## Rate limiting

The MCP endpoint is throttled per client IP address over a one-minute window. The budget is generous enough for an ordinary agent session — in stateless mode every tool call is its own request — but it is not unlimited.

When you exceed it, the endpoint answers `429 Too Many Requests` with a `Retry-After` header giving the number of seconds until the window resets. Honour it; retrying earlier only consumes the next window. Requests that arrive without a resolvable client IP share a single, much smaller budget, so send your requests through a proxy that sets `X-Real-IP` correctly, or none at all.

Plain HTTP requests for the pages, the Markdown versions and the discovery documents are not throttled by the application. Please still fetch at a reasonable pace and prefer `llms-full.txt` or the MCP server over crawling every page individually — both give you the same content in far fewer requests.

## Identifying yourself

Optional, and never a condition of access. A descriptive `User-Agent` is welcome and helps the publisher see which systems use the source; requests do not need to be signed, and unsigned requests are not treated differently.

## Contact and legal

- Imprint: https://www.sl-armaturen.com/impressum
- Privacy policy: https://www.sl-armaturen.com/datenschutz