What changed

Anthropic says message threads are in public beta on the Claude Platform and are used through Messages API requests, so an agent can stop resending the full history on every turn. A coding agent might make hundreds of requests in one session, and each request would carry every earlier message and tool result. Once a conversation reaches hundreds of thousands of tokens, that resend adds time before Claude starts responding.

The first request includes the full conversation and starts a thread. Later requests include only the new messages and the ID of the previous response. Between turns the thread stays in memory, encrypted, and not written to disk. It expires five minutes after the last request, and the client then resends the full conversation to start another.

Claude still reads the full conversation on every turn, so token usage stays the same. Standard Claude token rates and existing rate limits apply. ClaudeDevs posted on November 2, 2026, called that in-memory hold a server-side cache, and said to send the previous message ID. On each later turn, send the ID of the previous response, as Anthropic's article specifies.

Who it is for

Anthropic aims message threads at active sessions: long-running coding agents, computer-use agents whose screenshots stay with the thread instead of being re-uploaded on every turn, and document or image-heavy workflows. Anthropic says that more carried context can reduce time to first token further.

With message threads on by default for most Claude Code users, Anthropic reports that requests of 500,000 to 1 million tokens see median time to first token fall by about 35 percent. ClaudeDevs reported that same median. For long conversations, ClaudeDevs said the median request went from about 800 KB to 1.8 KB, while token usage stays the same. Anthropic's article says the requests are smaller. The 800 KB and 1.8 KB figures are from the ClaudeDevs post.

Public beta covers the native Claude Platform, Claude Platform on AWS, Claude in Microsoft Foundry, and Claude Models on Google Cloud (Anthropic managed).

How to try it

Add the thread parameter to Messages API requests, then follow Anthropic's documentation. ClaudeDevs also said to add a beta header.

On Claude Code, Anthropic says message threads are already on for most users.

Limits of the beta

No general-availability date is given, and Anthropic names no Claude Code version for the default. Who counts as most Claude Code users is not defined. The announcement does not include the thread parameter's fields, errors, or a beta-header name.

Source

Claude Developers · @ClaudeDevs

Also from

Sources and editorial approach

Sources checked on November 2, 2026: Anthropic's message threads article on claude.com, and the ClaudeDevs post at 2026-11-02T19:06:08Z. Claude Code changelog notes from version 2.1.316 (October 31, 2026) through version 2.1.300 (October 14, 2026) do not mention message threads. The rest of the 2.1.299 note and earlier entries on that page were not covered. The Messages API reference for the thread parameter, any beta header name, and error behavior was not covered. No separate latency or request-size measurement was run.