Anthropic Adds Message Threads to the Claude API

Anthropic introduced message threads in a company article, making the feature available in public beta through the Claude Messages API. Message threads let agents send a conversation once, then submit only new messages and the previous response ID on later turns.

Anthropic says the feature reduces time to first token during long sessions. For requests containing 500,000 to 1 million tokens, median time to first token fell by about 35% among most Claude Code users, where message threads are enabled by default. Claude still reads the full conversation on each turn, so token usage remains unchanged.

Threads are held in encrypted memory and are not written to disk. They expire five minutes after the last request, after which the client must resend the full conversation. The public beta is available on the native Claude Platform, AWS, Microsoft Foundry, and Anthropic managed Claude models on Google Cloud.