ONE CATALOGUE. TWO WAYS IN.
A directory your agent can read.
The same sourced records power these pages and the machine-readable catalogue. No screenshots or page scraping needed.
Full catalogue
All 10 services, capabilities, connection details and evidence dates.
/directory.jsonSCHEMARecord schema
Field types and meanings for reading and validating catalogue updates.
/directory.schema.jsonTXTAgent reading guide
A short index of the directory and individual records.
/llms.txtStart with a question
Find Icelandic weather services that only read information. Show the setup instructions and tell me what has actually been checked.
An agent can fetch the catalogue, match category and actions, and return the source-backed record. The JSON file is a complete snapshot; filters run in your client.
{
"category": "Weather",
"actions": "read",
"connectionTest": null
}Stable records and clear evidence
- Every service has a stable ID and an individual HTML, JSON and Markdown page.
lastSourceReviewrecords when documentation was checked. It does not describe software age or uptime.connectionTest: nullandindependentSecurityReview: nullmean that no such check is recorded.- A hosted URL or local package name points to the provider. Reading this directory does not connect you to that service.
Discovery is still a distribution task
Readable pages, a sitemap and structured records give search engines and agents useful entry points. A domain alone does not make every agent aware of the directory. Client integrations and links from relevant providers will also matter.
This first version provides a static JSON feed. It does not yet expose its own MCP server or claim compatibility with the full MCP Registry API. The reading guide is an additional entry point, not a universal discovery standard.
Public access: people and agents can read these pages and feeds without signing in. This address hosts a directory; it does not provide an MCP endpoint at /mcp.