Anthropic has introduced Message Threads in public beta on the Claude Platform, giving developers a way to send new conversation messages without uploading the entire history on every request.
Announced on November 2, the feature targets long-running agents. Anthropic says it is already enabled by default for most Claude Code users.
For requests containing between 500,000 and one million tokens, the company reports an approximately 35% reduction in median time to first token. That measures how quickly a response begins, rather than how long the complete task takes. The figure is Anthropic’s reported result, not an independent TechLayers benchmark.
How Message Threads Works
Developers start a thread by sending the complete conversation. Subsequent requests include new messages and the previous response’s ID, allowing the API to reconstruct the conversation.
Anthropic’s documentation specifies a thread parameter and the message-threads-2026-08-12 beta header for direct API requests.
| Stage | What happens |
|---|---|
| Create | Send the full conversation with thread.type set to create. |
| Continue | Send new messages with thread.type set to continue and the previous response ID. |
| Recover | If the thread becomes unavailable, resend the full conversation in a new create request. |
Threads expire after five minutes without another request. Developers must retain their own conversation history; the temporary thread does not replace application storage. Anthropic’s SDK helper can handle replay when a thread cannot be continued.
Smaller Requests Do Not Mean Lower Token Usage
Claude still reads the full conversation on each turn. Message Threads therefore does not, by itself, reduce token usage. Standard token pricing and existing API rate limits continue to apply.
The beta is available through the native Claude Platform, Claude Platform on AWS, Claude in Microsoft Foundry and Anthropic-managed Claude Models on Google Cloud.
Prompt caching remains a separate feature and is not automatically enabled by creating a thread. Anthropic’s documentation also lists Message Threads as ineligible for Zero Data Retention and HIPAA use, a compatibility detail organisations should check before adoption.
What the Update Means for Developers
The practical benefit is most relevant to agents that repeatedly exchange large amounts of context, such as coding assistants handling tool results or workflows involving documents and screenshots.
For teams evaluating the beta, the useful test is whether smaller requests improve responsiveness in their own application. Measure time to first token separately from total completion time, and test recovery after an inactive session.
The five-minute expiry also makes reliable conversation storage essential. A responsive active session still needs a working recovery path when its temporary thread disappears.
