The four tiers
Limits shown are defaults. Your tenant can configure per-tier overrides — see tier limits.
Tier 0: Registered
The starting tier. An agent is created under a parent account and immediately becomes Tier 0. No declarations, no history, no vouching — just “this agent exists.”- Safe for prototypes, dev, minimal-trust automations
- Spending capped low enough that a runaway loop is financially containable
- No additional docs required
Tier 1: Declared
The agent’s operator files a Declared Spending Declaration (DSD) — a self-attestation stating:- What the agent will be used for (natural-language purpose statement)
- Expected merchant categories
- Expected monthly volume
- Who’s accountable (a named individual in your org)
Tier 2: Verified
Two paths: Path A — Historical record. Agent has operated at Tier 1 for 30 days with no violations, disputes, or manual freezes. Automatic upgrade. Path B — Enterprise override. Your tenant’s operations team approves the upgrade based on their own due diligence. Useful for agents where the operator has its own established history (e.g. a treasury bot for a Fortune 500 customer).Tier 3: Trusted
The highest tier. Requires:- Kill-switch operator — a named human (not the agent itself) who can instantly revoke credentials. Their contact details are held in the agent record.
- Continuous Agent Integrity (CAI) — an ongoing signal that the agent code and behavior haven’t drifted. Usually wired up to your CI/CD + runtime monitoring.
Effective limits
Agent KYA tier is combined with the parent account’s KYC/KYB tier. The effective cap is the minimum of both:Read the current limits
Violations
If an agent attempts a spend over its effective limit, the API returns403 KYA_LIMIT_EXCEEDED. The attempt is logged. Repeated violations can trigger automatic tier downgrade (configurable per tenant).