Anthropic launches Message Threads beta to speed up long-running Claude agents
Announced · Published · By Contractor AI Shop
Anthropic released Message Threads in public beta, keeping long agent conversations on its servers between requests. Anthropic says this cut median time-to-first-token about 35% on 500k–1M-token requests in Claude Code usage, with token usage and pricing unchanged.
Key facts
- Anthropic says Message Threads is in public beta and available on the Claude Platform, Amazon Bedrock (AWS), Microsoft Foundry and Google Cloud. Anthropic
- Anthropic-stated: median time-to-first-token dropped about 35% on 500k–1M-token requests, based on its Claude Code usage. We haven't tested this. Anthropic
- Per Anthropic's docs, threads are encrypted in memory and expire 5 minutes after the last request; the feature is not eligible for Zero Data Retention. Claude Platform Docs
- Anthropic says token usage and pricing don't change; The New Stack notes the speedup doesn't lower your token bill. The New Stack (Paul Sawers)
- The Decoder frames the feature as aimed at long-running AI agents. The Decoder (Matthias Bastian)
Our take: what it means for contractors
This mainly matters if you or a developer run long Claude agent sessions, like Claude Code on big projects. Expect faster responses, not lower bills. If your contracts require zero data retention, skip it for now.
What we don't know yet
- When the beta ends or becomes generally available
- Speed gains on shorter requests or non-Claude Code workloads
- Independent benchmarks
Sources
- Anthropic · Make long-running agents faster with message threads · 2026-11-02 · Primary
- Claude Platform Docs · Message threads · 2026-11-02 · Primary
- The New Stack (Paul Sawers) · Anthropic speeds up Claude Code with message threads. But your token bill stays the same. · 2026-11-02 · Secondary
- The Decoder (Matthias Bastian) · Anthropic beschleunigt langlaufende KI-Agenten mit neuen "Message Threads" · 2026-11-02 · Secondary