Developers

API & MCP.

Real restock-cadence over the DropWatch archive. Free board, paid per-item prediction (key or x402).

REST

EndpointTierWhat
GET /nichesfreeNiche keys + labels.
GET /v1/board?niche=tcgfreeItems ranked by restock frequency.
GET /v1/predict?niche=tcg&query=...paid (key or x402)Next-restock-window prediction + confidence.
GET /healthfreeLiveness + archive depth.
curl "https://cadence.dropwatchhq.com/v1/board?niche=aqua"
# prediction via x402 (HTTP 402 with x402 challenge until paid):
curl "https://cadence.dropwatchhq.com/v1/predict?niche=aqua&query=anubias"

MCP

Endpoint https://cadence.dropwatchhq.com/mcp. Tools: list_niches, cadence_board (free), predict_restock (premium, x402). Discovery: /.well-known/x402, /llms.txt.

curl -X POST https://cadence.dropwatchhq.com/mcp -H 'content-type: application/json' \
 -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"cadence_board","arguments":{"niche":"aqua"}}}'
Confidence is honest: predictions carry HIGH/MEDIUM/LOW/INSUFFICIENT based on observed event count, history depth, and regularity. Thin items degrade to a tracker, not a guess.