# OPERANDI

> The operation layer for embodied robots and AI agents. Identify a real appliance,
> pull a grounded, safety-gated, machine-executable operation package: every control
> with a visual anchor, ordered procedures with per-step verification, force limits,
> hazards and interlocks — re-expressed from the appliance's own documentation,
> verified deterministically, never guessed at runtime.

Base API: https://api.operandi.cc (JSON; OpenAPI at /openapi.json; explorer at /docs)
INSTANT START (no signup): POST /v1/trial → {api_key} with 2 packages included. Use it at once.
MCP: `pip install operandi-mcp` — zero config; the server mints its own trial key on first use.
Auth: `Authorization: Bearer ok_live_...` or `X-API-Key`. Signup (free, 10 pkgs/month): POST /v1/auth/signup
Past any cap: 402 with X-Signup-Url / X-Upgrade-Url headers — the funnel is machine-readable.

## Core flow
- POST /v1/identify        — observed text (nameplate OCR, model string) → ranked matches
- GET  /v1/operate/{slug}  — the full robot-ready package in one call
- GET  /v1/objects         — browse/search the catalog
- GET  /v1/catalog/stats   — corpus size and validation-tier totals
- GET  /v1/demand          — request models we don't hold yet (feeds acquisition)

## Trust model
Packages carry validation_tier: auto → sim_validated → phys_validated → robot_cleared.
A deterministic groundcheck verifies claimed numbers against source documents (~96% clean).
Unknown model = honest 404 + demand hook, never a hallucinated procedure.

## Docs
- Human docs: https://operandi.cc/docs/
- MCP server: operandi-mcp (packaged, publication pending)
- Contact: https://wa.me/353874348585