Spending Policies

Spending policies define guardrails for agent transactions. Every payment passes through the policy engine before reaching compliance screening or on-chain submission. Policies are evaluated in Redis with sub-10ms latency.

Rule Types

Rule	Field	Description
Per-Transaction Limit	`max_per_transaction`	Maximum amount for a single payment
Daily Limit	`max_daily`	Maximum cumulative per 24-hour rolling window
Weekly Limit	`max_weekly`	Maximum cumulative per 7-day rolling window
Monthly Limit	`max_monthly`	Maximum cumulative per 30-day rolling window
Counterparty Whitelist	`counterparty_whitelist`	Only these addresses can receive payments
Counterparty Blacklist	`counterparty_blacklist`	Payments to these addresses are rejected
Velocity Limits	`velocity_limits`	Max transactions per hour, max unique counterparties per day

Advanced Rules

Rule	Field	Description
Allowed Chains	`allowed_chains`	Restrict to specific chains
Allowed Stablecoins	`allowed_stablecoins`	Restrict to USDC, USDT, or EURC
Category Restrictions	`category_restrictions`	Allow or block payment categories
Time Restrictions	`time_restrictions`	Limit to specific hours, block days
Escalation	`escalation_rules`	Require human approval above threshold

Enforcement Modes

Mode	Behavior
`enforce`	Violations blocked. Payment rejected with `POLICY_VIOLATED`.
`warn`	Violations logged but payment proceeds. Response includes `policy_result: "violated"`.
`audit_only`	Violations silently logged. No impact on payment.

Violation Actions

Action	Behavior
`reject`	Payment rejected (HTTP 403)
`flag`	Payment proceeds, flagged for review
`escalate`	Payment held for human approval (HTTP 202)
`suspend_agent`	Payment rejected and agent suspended

Evaluation Flow

Step	Action	Latency
1	Load active policies from Redis cache	<1ms
2	Sort by priority (lower = higher priority)	<1ms
3	Evaluate each rule against the transaction	<5ms
4	Aggregate: if any enforced rule fails, violation	<1ms
5	Write to `oris_policy_evaluations` hypertable	async

SDK Usage

Python

policy = agent.set_policy(
    max_per_tx=50,
    max_daily=500,
    max_weekly=2000,
    counterparty_whitelist=["0xabc...", "0xdef..."],
    enforcement_mode="enforce"
)

# Simulate before sending
sim = agent.simulate_payment(amount=75)
print(sim.passed)          # False
print(sim.violated_rules)  # ["max_per_transaction"]
print(sim.verdict)          # "reject"

See Policies API for endpoint details.