[{"data":1,"prerenderedAt":14},["ShallowReactive",2],{"artikel-en-berita-anthropic-message-threads-api-2-november-2026":3},{"slug":4,"lang":5,"title":6,"description":7,"date":8,"updated":9,"topic":10,"product":9,"cover":11,"readMin":12,"html":13},"berita-anthropic-message-threads-api-2-november-2026","en","Anthropic Launches Message Threads: AI Agents No Longer Resend the Whole Chat","Anthropic brings message threads to the Claude API in public beta, cutting the wait for a first reply on long agent sessions by about 35 percent.","2026-11-03",null,"berita","\u002Fimages\u002Fartikel\u002Fberita-anthropic-message-threads-api-2-november-2026.webp",3,"\u003Cp>If you have ever built an AI agent, you know the problem: every time the agent makes a new request, it resends the entire conversation history. On long sessions that means hundreds of thousands of tokens sent over and over, and every request gets slower. On November 2, 2026, Anthropic announced a fix called message threads.\u003C\u002Fp>\n\u003Ch2>1. What message threads are\u003C\u002Fh2>\n\u003Cp>Message threads are now in public beta on the Claude Messages API. With this feature, an agent sends the full conversation once. After that, each request carries only the new messages, plus the ID of the previous reply.\u003C\u002Fp>\n\u003Cp>Anthropic's servers keep the conversation, so requests stay small even as the session grows. The official release notes say a thread expires five minutes after its last request and is deleted no more than an hour after its last use.\u003C\u002Fp>\n\u003Ch2>2. How much faster it gets\u003C\u002Fh2>\n\u003Cp>Anthropic says that on requests of 500,000 to 1 million tokens, the median time to first token drops by about 35 percent. The figure comes from Claude Code users, where threads are already on by default for most people.\u003C\u002Fp>\n\u003Cp>The company also shared another number: in long conversations, the median request size fell from about 800 KB to 1.8 KB. That is a large difference for agent sessions that make hundreds of requests.\u003C\u002Fp>\n\u003Ch2>3. It saves time, not money\u003C\u002Fh2>\n\u003Cp>One thing to understand clearly: Claude still reads the whole conversation on every turn. Anthropic confirms token prices and rate limits stay the same. Your bill does not go down, replies just arrive sooner.\u003C\u002Fp>\n\u003Cp>For cost savings, Anthropic took a different route: it cut cache prices by 75 percent in September, because long agent sessions are exactly what drives AI bills up.\u003C\u002Fp>\n\u003Ch2>4. How the data is stored\u003C\u002Fh2>\n\u003Cp>Threads are held in memory, encrypted, and never written to disk. Once a thread expires, the client has to resend the full conversation to start a new one.\u003C\u002Fp>\n\u003Cp>Anthropic also states that threads are not eligible for its zero data retention or HIPAA health-data setups. Developers with strict compliance needs should take note.\u003C\u002Fp>\n\u003Ch2>5. Where it is available\u003C\u002Fh2>\n\u003Cp>Message threads work with every Claude model across four channels: the Claude Platform, Claude Platform on AWS, Claude Models on Google Cloud, and Claude in Microsoft Foundry. Standard token rates and existing rate limits apply.\u003C\u002Fp>\n\u003Cp>The feature fits active sessions where latency matters: long coding agents, computer use agents that no longer re-upload screenshots every turn, and document or image-heavy workflows.\u003C\u002Fp>\n\u003Ch2>What it means for you\u003C\u002Fh2>\n\u003Cp>If you are a regular ChatGPT or Claude user, there is nothing for you to do. But if you pay for agent tooling or use Claude Code, this means your tools get more responsive at no extra cost. It is also a signal of where the industry is heading: the contest between AI providers is no longer only about which model is smartest, it is about long work sessions that stay fast.\u003C\u002Fp>\n\u003Col>\n\u003Cli>\u003Cstrong>Claude Code users:\u003C\u002Fstrong> threads are already on by default for most users, so long sessions should feel snappier.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Agent developers:\u003C\u002Fstrong> try the beta if your product runs long sessions, and test latency on your own workload before shipping.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>AI service buyers:\u003C\u002Fstrong> ask your vendor whether they have adopted threads, because they can cut waiting time without raising costs.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Cost watchers:\u003C\u002Fstrong> remember this does not lower your bill, so keep relying on caching and model choice to save money.\u003C\u002Fli>\n\u003C\u002Fol>\n\u003Ch2>Sources\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fclaude.com\u002Fresources\u002Farticles\u002Fmessage-threads\" target=\"_blank\" rel=\"noopener\">Make long-running agents faster with message threads\u003C\u002Fa> (Anthropic)\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Frelease-notes\u002Fapi\" target=\"_blank\" rel=\"noopener\">Claude Platform release notes, November 2, 2026\u003C\u002Fa> (Anthropic Docs)\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fthenextweb.com\u002Fnews\u002Fanthropic-message-threads-beta-claude-api-agents\" target=\"_blank\" rel=\"noopener\">Anthropic's message threads let Claude agents skip resending chats\u003C\u002Fa> (The Next Web)\u003C\u002Fli>\n\u003C\u002Ful>\n",1793685874793]