# ProductWatch — product compatibility registry

> A continuously updated registry of product models and their compatibility
> attributes, built for AI agents. Resolve a messy consumer string
> ("does this fit my fenix7 pro?") to ranked canonical models with lug width,
> attachment system, case size, confidence and source URLs.

## Access

- REST: `GET /api/public/v1/{dataset}/search?q={text}&limit=10` with header `X-API-Key: <key>`
- REST: `GET /api/public/v1/{dataset}/models/{slug}`
- REST: `GET /api/public/v1/datasets`
- OpenAPI: `/api/public/v1/openapi`
- MCP (Streamable HTTP, no key): `/mcp` — tools `find_watch`, `get_watch`, `list_datasets`

## Datasets

- `watches` — Garmin and Apple Watch models. Attributes: `lug_width_mm`,
  `attachment_system` (quickfit | quick_release | spring_bar | apple_slide | screw_in),
  `case_size_mm`, `band_size_family`.

## How to use the results

1. Send the consumer's raw text as `q`. Do not pre-clean it.
2. Read `ambiguous`. When true, ask one clarifying question instead of guessing.
3. Match on `fitments[]` where `fit = "native"`; `fit = "compatible"` means it works
   but is not the designed mechanism.
4. Never answer compatibility from model memory. If the registry has no confident
   match, say so.

## Human pages

- /docs — API and MCP documentation
- /datasets/watches — browsable table of every model