Skip to main content
Saturn is the execution layer for AI agents. It sits between your agent and the APIs it calls — handling auth, routing, budget enforcement, and per-call receipts.

Quickstart

Deploy your first agent in 5 minutes

Capabilities

10 capability verbs for AI agents

SDK Reference

TypeScript SDK documentation

Pricing

Pay-as-you-go with prepaid credits

​
What Saturn Does

  • Routes requests to 15+ upstream providers (OpenAI, Anthropic, Serper, E2B, and more)
  • Enforces budgets with per-agent caps and daily limits
  • Processes payments via card (USD) or Lightning (sats)
  • Logs everything with full audit trail and cost attribution

​
Runtime Enforcement vs Dashboards

Dashboards show you what happened. Saturn prevents what shouldn’t happen.
ApproachWhen it actsOutcome
Dashboard alertsAfter spend occursYou see the bill
Saturn enforcementBefore upstream callCall rejected if over budget
Saturn operates at the execution boundary. Every capability call passes through policy checks before reaching any upstream provider. Over-budget calls are rejected—not logged, not alerted, rejected.

​
Architecture

┌─────────────────────────────────────────────────────────────┐
│                        Your Agent                           │
└─────────────────────────────┬───────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                         Saturn                              │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────────────┐  │
│  │   Policy    │  │   Ledger    │  │   Provider Router   │  │
│  │   Engine    │──│   (Credits) │──│   (OpenAI, etc.)    │  │
│  └─────────────┘  └─────────────┘  └─────────────────────┘  │
└─────────────────────────────┬───────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│              Upstream Providers (OpenAI, Anthropic, etc.)   │
└─────────────────────────────────────────────────────────────┘
Every call:
  1. Authenticated via agent API key
  2. Checked against policy (caps, allowed capabilities)
  3. Quoted against current credit balance
  4. Executed only if all checks pass
  5. Settled atomically with receipt issued

​
When to Use Saturn

Autonomous Agents

AI agents that call paid APIs without human oversight

Multi-tenant SaaS

Per-user AI features with isolated budgets

Production Systems

Where cost overruns are unacceptable

Multi-Provider

Unified budgets across OpenAI, Anthropic, and more

​
When Not to Use Saturn

  • Single-user CLI tools with manual oversight
  • Development environments where cost is irrelevant
  • Systems where you want unbounded spend
  • Scenarios requiring direct provider API access for unsupported features