Sui Agent Payments x Alibaba Cloud: Per-Call USDC Billing Is Live — But Who Scores the Call?
Sui Agent Payments × Alibaba Cloud: Per-Call USDC Billing Is Live — But Who Scores the Call? On October 7, 2026 — announced live at Sui Basecamp — Sui Foundation and Alibaba Cloud made cloud + AI services available through Sui Agent Payments. The shape: the user first hands the agent an expenditure account with preset limits (which services, total budget, max per transaction, expiry time). Then…
On October 7, 2026, the Sui Foundation and Alibaba Cloud unveiled a new service through Sui Agent Payments. The service allows users to hand agents an expenditure account with preset limits, including the services allowed, total budget, maximum per transaction, and expiry time. When the agent makes a service call, it is charged in USDC on the Sui network.
The pricing information is available through SKILL.md and JSON endpoints, with an example being a $0.001 service costing exactly $0.001 per call. The Sui Agent Payments system supports MCP (Multi-Channel Payment) and registers services as tools in Claude Code and Codex, with sponsors covering the gas costs. An account acts as a gate to restrict the agent's spending to allowed services, with a total budget and max per transaction as the caps.
The expiry time is a time-bound authority lease that has not been announced. While the account sets the boundaries for spending, it does not determine whether the agent should spend. During a live test on October 8, 2026, two Sui-shaped instructions were scored against a real decision gate with bands of 0.80 and 0.50. The first call to a $0.001 Alibaba Cloud OCR service matched the SKILL.md catalog, was within the $10 budget, and had a per-transaction cap of $1.
The service was listed on the allowlist, resulting in a score of 0.35, with the option to escalate or block and log. The second call was to a $4,200 unlisted GPU training job, which was not in the catalog, and the account was about to expire in 10 minutes. A proposed 10,000-call loop by the agent itself was not cleared by the user, resulting in a score of 0.35, with escalation and blocking and logging as the options.
The judgment, not the friction, is the upgrade in this system, as a calibrated gate auto-runs catalog calls with a score of ≥0.80 and escalates unlisted loops with a score of 0.50. All receipt tables, the VAA accountability half connection, the 5-step per-call gating DIY, and the exact curl commands can be found on the canonical page: https://scriptmasterlabs.com/sui-agent-payments-alibaba-cloud.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.