Anthropic launches Claude Agents SDK to general availability

Read the original article →

What happened

Anthropic moved its Claude Agents SDK to general availability on August 5. The SDK includes orchestration primitives for multi-step agents, built-in tool routing, human-in-the-loop checkpoints and persistent session memory across turns.

It integrates with Claude Sonnet 5 and Opus 4 and ships reference architectures for customer support, code review and research workflows. Pricing follows standard Claude API token rates with no separate platform fee at launch.

Why it matters

Every major lab now offers an agent framework, but Anthropic's GA signals it expects production deployments, not just demos. Checkpoints and memory are the features enterprises ask for once agents run longer than a single prompt.

Reference architectures lower the time to a working prototype, which accelerates internal pilots.

MintedBrain take

Start with one reference workflow close to your use case and instrument token use from day one. Agent SDKs make it easy to chain steps; spend controls and human checkpoints keep costs and errors bounded.

References

This article was originally published at Anthropic Docs. For the full piece, read the original article.

Discussion

  • Loading…

← Back to News