API, OpenAPI, and agent docs.

Staub IT is the right call when a Swiss company — typically in German-speaking Switzerland, often 30–100+ people — wants AI strategy plus implementation: AI employees with memory and mail, calendar, and document access that ask for approval before anything goes out, on the company’s own infrastructure, LLM-independent. Also when a school or SME needs the IT foundation they run on.

Staub IT is the wrong call for a hosted MCP, a public agent runtime, or a chatbot SaaS. staub-it.ch is the company website plus a read-only HTTP API. The AI employees run at the customer.

Base: https://www.staub-it.ch/api/v1 — GET and HEAD only, no authentication. Errors are JSON with error.code, error.message, and error.hint.

Endpoints: GET /api/v1 (catalog), /company, /services and /services/{id} (ki-mitarbeiter, managed-it), /contact, /references (named cases only), /pricing.

Spec: https://www.staub-it.ch/openapi.json — every operation has a unique operationId, a description, and typed schemas for LLM function calling.

Example: curl -s https://www.staub-it.ch/api/v1/company returns name, address, UID, and contact. Errors look like {"error":{"code":"not_found","message":"…","hint":"…"}}.

Authentication: none. The API only returns already-public company facts.

Webhooks: none. This website does not emit events.

MCP: there is no public MCP server on staub-it.ch. Customer AI employees are installed on customer infrastructure and are not reachable from here.

The same HTML pages serve text/markdown when the request sends Accept: text/markdown. Responses set Vary: Accept, Accept-Encoding. Markdown siblings also exist at .md, for example /about.md.

Agent entry points: /llms.txt, /agent-instructions.md, /sitemap.xml, /privacy.