Anthropic launches Message Threads in public beta so long-running agents send only new turns, cutting time to first token by about 35%
Anthropic has released Message Threads in public beta on the Claude Platform. Until now every Messages API request had to resend the full conversation history; with a thread, an agent sends only the messages added since the last turn. Anthropic says the feature is on by default for most Claude Code users and that requests of 500,000 to 1 million tokens see median time to first token (TTFT) fall by about 35%, with token usage and pricing unchanged.
The first request sends the full conversation, system prompt and tools as usual and adds a thread parameter to create the thread; each later request sends only the new messages and the previous response’s id, and the API rebuilds the conversation server-side. According to the documentation, requests need the message-threads-2026-08-12 beta header; system and tools can only be sent on the create request, and fixed fields such as the model must stay the same on every turn or the API returns a 400 error.
Threads are short-lived. Anthropic says they are held encrypted in memory and not written to disk, and expire 5 minutes after the last request; the documentation says they are deleted no more than 1 hour after last use. A continue request on an expired thread returns a 404, and the client must resend the full conversation to start a new one, so the docs advise applications to keep their own copy of the conversation; the SDK thread helper handles this replay.
Because Claude still reads the full conversation on every turn, token usage is unchanged and standard token rates and existing rate limits apply. Anthropic says the biggest gains are for long-running coding agents, computer-use agents whose screenshots stay with the thread, and document- and image-heavy workflows. The feature is available on the native Claude Platform, Claude Platform on AWS, Claude in Microsoft Foundry and Claude Models on Google Cloud (Anthropic managed).
Sources
- Make long-running agents faster with message threads — Anthropic（Claude）
- Message threads — Anthropic（Claude Platform 文档）
- Anthropic’s Message Threads cut time to first token by 35 percent for long conversations — The Decoder