Claude message threads shrink later API turns

Anthropic's platform release notes for 2 November 2026 add message threads to the Messages API, in beta with the message-threads-2026-08-12 header. The message threads guide (opens in a new tab) says you send the full conversation once, with the system prompt and tools, using thread set to {"type": "create"}. Each later turn sends only the new messages and the id of the previous response, with thread set to {"type": "continue", "previous_message_id": "..."}. The API rebuilds the conversation before it calls the model.

A thread is best-effort. It expires 5 minutes after its last request, and the API deletes it no more than 1 hour after its last use. If a continue request returns 404, you send the full conversation again. Your own copy of the messages stays the source of truth. Threads are not eligible for zero data retention. The guide says a thread is held in memory, encrypted, and never written to durable storage. Threads do not change pricing or rate limits. They are in beta on every model, on the Claude API, Claude Platform on AWS, Claude Models on Google Cloud (Anthropic managed), and Claude in Microsoft Foundry.

Anthropic's announcement (opens in a new tab) says message threads are on by default for most Claude Code users, and that requests of 500,000 to 1 million tokens see a median time to first token drop of about 35%. The same note says Claude still reads the full conversation, so token usage stays the same.

Who it's for: a developer whose Messages API agent resends a long history on every turn.

Fits: Best for Vibe Coding stays Cursor for an edit you can reject in the project. Claude Code stays second, for a job you hand off from the terminal. A shorter wait before the first token does not change that split. The coding guide, Claude Code alternatives, and Cursor vs Claude Code are unchanged. The verdict is unchanged.