Examples
Practical patterns for common agent scenarios
Spending Cap on a Shopping Agent
Issue a passport with a $100/day spend limit. The enforcement layer blocks any purchase that would exceed the cap and produces a signed attestation for every decision.
Gate-Protected API
Put a gate in front of your API. Credentialed agents get full access, anonymous agents get read-only with rate limits, and invalid passports are denied outright.
Human-in-the-Loop Approvals
Configure an approval threshold so any action over $200 pauses the agent and sends you a text. Reply YES or NO from your phone.
Domain-Restricted Agent
Lock an agent to only call specific domains — your internal tools and nothing else. Every blocked request is logged with a signed attestation.
MCP Server with Modus
Connect Modus to Claude Desktop in one command. Your agent gets a passport, and every tool call is verified and logged.
Agent-to-Agent Commerce
Agents discover services, compare signed pricing catalogs, and transact with locked terms. Every call produces a metered receipt for settlement.
Multi-Level Security Profiles
Configure gates with L1 (basic), L2 (trusted issuers), or L3 (strict + proof-of-possession) security. Higher tiers require stronger agent credentials.
Guardrail Templates
Use pre-built constraint templates for common agent types — customer support, research, purchasing — and customize from there.
For step-by-step setup, see the guide. For the full reference, see the documentation.