Message Threads

Anthropic's beta Messages API feature that keeps conversation state server-side so clients send only new messages.

Anthropic · Infrastructure and inference/serving · tool

Published in Feed

Overview

Anthropic opened Message Threads as a public beta of the Messages API on November 2, 2026. After one full-conversation request, clients send only new messages while the server keeps state between turns. Threads are best-effort and expire five minutes after the last request, and a lost thread returns 404 thread_not_found, so apps must keep their own copy of the conversation. Anthropic reports about 35% lower median time to first token for requests of 500K to 1M tokens, and threads are on by default for most Claude Code users.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Send only new messages after the first full request
  • Beta header message-threads-2026-08-12; SDK helper client.beta.messages.thread()
  • Threads expire five minutes after last request; 404 thread_not_found on loss
  • About 35% lower median time to first token for 500K-1M token requests (Anthropic-reported)
  • Not supported with Message Batches, token counting or most context editing; unavailable with Inference hooks
  • Available on Claude Platform, Claude Platform on AWS, Claude in Microsoft Foundry and Claude Models on Google Cloud

Use cases

  • Long-running coding or research agents with very large contexts
  • Cutting request payload size for chatty multi-turn clients

Pricing

Billed per token at standard rates; billing still counts the full context processed.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
2 Nov 2026
Record updated
7 Nov 2026
Record last checked
7 Nov 2026

Explore the Feed

Other entries filed under infrastructure and inference/serving; inclusion does not imply equivalent features.

Browse the Feed