ZBS Index What actually exists in applied AI, with the source next to it

A catalog written for agents to read

The reader this site is built for is an AI agent choosing a tool, not a person browsing. Every page is static HTML with no JavaScript, every claim carries its source and observation date in the markup, and the whole catalog is available as JSON at /api/v1/catalog.json. People are welcome; the format is not optimised for them.

We track 23,822 entries and have checked 10,523 of them against the repository their maintainer listed.

Of the 7,089 MCP servers we checked, 19% point at a repository that has been deleted or archived. The registry still lists them.

That number is why this catalog exists. Directories in this category compete on how many tools they list, and none of them tell you what share of those listings still point at something real.

Nothing here is scored. Every number carries what it measures, the window it covers, its source and the date it was observed. Where we have not checked something, the page says so instead of leaving you to assume.

Start from the job, not from the tool

Each of these is a reviewed starting stack for one task in one industry: the primary path, a private path and a budget path, with the limit and the source of every choice.

Solution stacks

Browse the catalog

Browse by capability

How to read anything on this site

A star is a bookmark somebody left once and never took back. A download count includes continuous integration and mirrors. Website traffic measures a company’s marketing site, not use of its product. None of those are quality, and we never add them together.

Recent maintenance says somebody is still tending the project. It does not prove quality or fit for your task, so it stays one visible signal rather than becoming a score.

Machine access

The whole catalog as JSON: every entry with its slug, version, capabilities, repository status and the date each field was observed. No key, no rate limit, no HTML parsing.

GET /api/v1/catalog.json