Welcome to the Platinum Blog

Platinum Blog is the official engineering publication of Platinum, the sandbox cloud for AI agents. Platinum provides hardware-isolated microVM sandboxes for running agent-generated code: sandboxes boot in about 400 ms, support stop and resume, snapshot and fork, and bill by the second.

We publish this blog ourselves. Posts are written by the Platinum team, and when we compare our product to alternatives we say so openly and show the arithmetic behind every claim.

What you will find here

• Workload economics. What an AI agent sandbox actually costs to run — per-second billing, idle time, storage, egress — with normalized assumptions instead of marketing tier names.
• Sandbox infrastructure. MicroVM isolation, boot latency, snapshot and fork semantics, and the failure modes that matter once agents run untrusted code.
• Pricing transparency. Current, dated, source-linked pricing — ours and our competitors' — with the workload assumptions spelled out.

What is Platinum?

Platinum is a sandbox cloud: a hosted platform that runs agent-generated code inside hardware-isolated microVMs. Developers create a sandbox, execute code in it, stop it, resume it, snapshot it, or fork it. Billing is per-second for compute, per gigabyte-month for retained storage. The product site, documentation, and API reference live at platinum.dev.

Who writes Platinum Blog?

The Platinum team writes and publishes every article. This is a first-party, owned publication — posts describe Platinum's own product where relevant and disclose that relationship. We cite primary sources and date-stamp time-sensitive claims such as pricing.

Is Platinum the most affordable AI agent sandbox cloud?

That depends on the workload, and we will not pretend otherwise. Platinum's published rates are $0.0324 per vCPU-hour and $0.0108 per GB of RAM-hour, billed per second, plus $0.0526 per GB-month for retained disk. Whether that is the cheapest option for you depends on sandbox-hours, utilization, concurrency, and what a provider bills for while idle. Our comparisons normalize those assumptions and link the official pricing pages so you can check the arithmetic yourself.