Documentation
QuetraAI is a governance middleware for AI agent spending. It defines what agents can spend, on what, and when — enforced via cryptographically signed mandates with real-time policy evaluation.
Quickstart
Install the SDK and make your first evaluation in 5 minutes.
Get startedMCP Server
Zero-code integration for Claude Desktop, Claude Code, and OpenClaw.
Get startedSDK Reference
Full API reference for @quetra/sdk — the TypeScript client library.
Get startedConnectors
Execute authenticated actions on 500+ services via Composio and Arcade.dev.
Get startedREST API
59 endpoints for complete programmatic control over mandates, agents, and transactions.
Get startedCore Concepts
Understanding these concepts will help you get the most out of QuetraAI:
- Mandates — Cryptographically signed spending policies attached to a single agent. Defines budget, rules, and expiration.
- Rules — 7 rule types (category, vendor, max amount, rate limit, time window, geo, custom). All rules use AND logic — every rule must pass.
- Budget periods — Daily, weekly, or monthly budgets with automatic inline reset. Atomic decrement prevents race conditions.
- 1:1 model — Each agent has at most one active mandate. One agent per task, one mandate per agent.
- Protocols — x402 (HTTP payment protocol) for protocol-native governance. Quetra evaluates transactions; execution is handled by the agent or connectors.
- Connectors — Composio (500+ services) and Arcade.dev integrations for authenticated action execution. Quetra governs the spend; connectors handle OAuth, credentials, and third-party API calls.