Daily · · Claude API · 🌶️Nice to knowMild: nice to know Why: Smaller requests in long agent loops
Claude API message threads send only new turns
Long agent loops can stop re-uploading the whole conversation every turn.
Message threads, in beta on the Claude Messages API since Nov 2, 2026, keep the conversation on the server so each later request carries only the new messages, with a lower time to first token.
What changed
Each follow-up request now carries just the new turn. Without a thread, every turn sends the full conversation again and the API validates and tokenizes all of it again.
With a thread, most of that work covers only the new turn, and Anthropic says time to first token stays lower and steadier as the conversation grows.
How to use it
- Send the
message-threads-2026-08-12beta header on every request in the thread. - On the first request, send the full conversation, system prompt and tools as usual, and add
"thread": {"type": "create"}. - On each later turn, send only the new messages with
"thread": {"type": "continue", "previous_, and leave system and tools out.message_ id": "<previous id>"}
The SDKs also have a thread helper that builds these requests and adds the beta header for you.
Keep your own copy
Threads are short-lived. One expires 5 minutes after its last request, and the API deletes it no more than 1 hour after its last use. When a thread can’t continue, the request returns a 404 error, and you resend the full conversation in a new create request.
That’s why the docs say to keep your own copy of every message. Threads work on the Claude API, Claude Platform on AWS, Claude Models on Google Cloud (Anthropic managed) and Claude in Microsoft Foundry. They aren’t eligible for zero data retention (ZDR) or HIPAA.
Source: Claude Platform release notes, Nov 2, 2026 · Claude docs: how message threads work