Pink data: 3 AI agents, 40 concurrent payments, one $6,000 daily limit. Did it hold?

说明与用途

测试表明，来自三个 AI 代理的并发支付请求被共享每日上限 $6,000 正确限制，未超过限额。

问题

Determine whether concurrent payment requests from multiple AI agents can exceed a shared company daily spending ceiling in the Pink Agentic AI Payments system.

结果

In all three runs, the total allowed payments matched the server’s company_spent_today and never exceeded the $6,000 ceiling. Blocked requests consistently cited the shared ceiling rule.

工作流

• 1. Set up sandbox with coffee template (3 agents, $6,000 company_daily_ceiling).
• 2. Configure agents a_purch, a_inv, a_pay with high monthly budgets and per-payment caps to isolate the shared ceiling rule.
• 3. Fire 40 POST /v1/payments requests concurrently from all agents using a single thread pool with no delay.
• 4. Record client-side responses: count allowed vs blocked and sum of allowed amounts.
• 5. Query server via GET /v1/budget to obtain company_spent_today.
• 6. Verify that client allowed sum equals server company_spent_today and does not exceed the ceiling.
• 7. Inspect blocked responses for rule and why fields to confirm shared ceiling enforcement.
• 8. Repeat the test three times (two with day reset, one fresh workspace).
• 9. Use public script 11-shared-ceiling-test for reproducibility.

证据

In every run, the client-side allowed sum matched the server's own company_spent_today exactly, and the total never went over the $6,000 ceiling.

来源

Every single blocked request across all three runs cited the shared ceiling specifically, not a per-agent rule.

来源

The why text shows a running total that increments across requests from different agents (a_purch, a_inv, a_pay all draw against the same counter).

来源