A container is a brand or product grouping over capabilities. One container, many capabilities — usually one per HTTP endpoint of that brand.
Without containers:
…appears in the registry as four independent rows. Agents have to mentally cluster them by slug prefix.
With a container:
…agents see one branded entity with a description, logo, website, and a verified domain badge. Discovery improves, abuse drops.
claimed_domain + domain_verified_at are present once the owner has proven control via DNS (see Domain verification).
Containers are optional — ungrouped capabilities still work fine and remain discoverable via GET /v1/capabilities.
The strongest signal a container’s owner truly represents a brand is DNS control: only dimenzuri.example.com’s owner can create a TXT record under that domain.
dimenzuri.example.com but slugs called pizza.delivery. A future hardening pass will require the slug’s domain segment to match.For agents using the Model Context Protocol:
Agents that want to find capabilities by brand should prefer list_containers → get_container → execute_capability over the flat list.