How can users allocate Anthropic Claude Agent credits?

Anthropic enables paid Claude Agent SDK credits, with a specific use window

Anthropic introduced a new way to buy access to Claude’s agentic tooling: paid Claude plans will include credits usable with its Claude Agent SDK. Customers can then allocate those credits for programmatic use of third-party agents.

The announcement also ties directly to how autonomous “agent” products plug into the Claude ecosystem. In practice, that means customers are no longer limited to using Claude only through a chat interface; instead, they can fund usage of an SDK that supports running external agents in a Claude subscription context.

A key operational detail is timing. Anthropic said the credit feature for paid plans becomes available starting June 15. Until that date, users would not have the new credit allocation mechanism.

For example, Anthropic specifically referenced third-party agents such as OpenClaw as an example of the kind of agentic software users could run through the SDK credit model.

Separately, the broader theme across related Anthropic updates is that the company is continuing to adjust the boundary between what is included with a Claude subscription and what must be accessed through third-party systems. The “credits” approach suggests Anthropic is monetizing agentic programmatic execution while still allowing interoperability with the wider agent developer ecosystem.

Why it matters: enterprises evaluating agent platforms care about (1) how agent tools are integrated, and (2) whether usage is trackable and billable in a straightforward way. Agent SDK credits are a step toward making agent execution feel more like a priced developer capability rather than an unpredictable, chat-only feature.