idlidu mlip — MLIP fitness oracle

idlidu mlip

Ensemble disagreement and trust bands on your structures — not another Matbench table. Submit a cell + property; get inter-model spread and a calibrated trust band.

Not a property-accuracy guarantee. Disagreement is a fitness-for-purpose signal. Structures are not retained by default. Cloudflare serves API/billing/docs only — no GPU on CF.

API surface (v1)

Host: mlip-api.idlidu.com · MCP: mlip-mcp.idlidu.com

POST /v1/fitness — structure → trust band + disagreement
GET /v1/ensemble · GET /v1/coverage
GET /v1/health · GET /v1/usage

Read the methodology →