Anthropic Opens Claude Message Threads Beta for Long Agent Sessions

Anthropic has opened message threads in public beta on Claude Platform, allowing developers to continue long-running agent conversations without resending the full history with every API request.

Median time to first token (TTFT) was about 35% lower for requests containing 500,000 to 1 million tokens. The feature is enabled by default for most Claude Code users.

The first request includes the full conversation and starts a thread. Each later request can include only new messages and the ID of the previous response. Claude still reads the full conversation on each turn, so the change does not reduce token usage.

Threads are encrypted and kept in memory rather than written to disk. They expire five minutes after the last request; after that, a client must send the full conversation to start a new thread.

The beta is also available on Claude Platform on AWS, Claude in Microsoft Foundry and Claude Models on Google Cloud. The feature is best suited to active sessions, including long-running coding and computer-use agents and workflows involving documents or images. Standard Claude token rates and existing rate limits apply.