Quartermaster API
AI procurement for restaurants, with deterministic spend controls (TrustRail) and PayPal sandbox payments
Quartermaster is a procurement agent for restaurants. An owner writes "We're expecting a busy weekend. Restock takeaway containers and cooking oil without exceeding $300." The agent forecasts demand, compares suppliers with Channel3 market prices and proposes carts. TrustRail checks every cart in code against the owner's Spend Mandate, blocking hidden fees and prompt-injected listings. PayPal then captures payment only after the owner approves one exact, signed cart.
This portal, the TypeScript and Python SDKs and the agent-friendly docs are generated by APIMatic from the API's OpenAPI 3.1 spec. A contract test checks that spec against the API's real responses.
Quickstart
Sign in, read stock and the Spend Mandate, and ask the agent to restock: in about 30 lines of TypeScript.
SDKs
Typed TypeScript and Python clients, with the session cookie and server environments built in.
API Reference
Agent-Friendly Docs
llms.txt and Markdown pages, so coding agents can integrate correctly.
Conventions
- Money is always integer minor units (cents): fields end in
_minor. - Sign-in sets an httpOnly session cookie:
__Host-qm_sessionon the hosted demo,qm_sessionlocally. - Every mutation sends
X-Requested-With: quartermaster(CSRF guard). Approval and payment calls also take anIdempotency-Key, so a retry never pays twice. - Errors share one shape,
ApiError, with a stablecodeand a fixed message. - Roles: owners approve and pay, managers request and research, viewers read.
- Payments run in the PayPal sandbox only. No real money moves.
Try it
The hosted demo is at quartermaster-web-o0df.onrender.com. Create an account with sample data, or use the shared sample workspace listed in the README.