LangChain: Anthropic Replay Fixes

Two Anthropic fixes target conversation replay failures. Merged PR 40864 serializes invalid tool calls so tool results keep matching parents, while open PR 40870 preserves nested caller context for server tools.

Duration: PT2M5S

Episode overview

This episode is a short developer briefing from LangChain.

It explains recent repository work in plain language.

  • Show: LangChain
  • Published: 2026-09-28T13:04:12Z
  • Audio duration: PT2M5S

Transcript excerpt

This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.

Good morning, it's Monday, September 28th, 2026, and this is your LangChain developer briefing.

The signal this cycle is Anthropic conversation replay — two focused fixes aimed at stopping valid follow-up turns from being rejected by the API.

First, handling malformed tool calls. A merged fix in PR 40864 changes how invalid tool calls on an AI message are sent back to Anthropic. Instead of dropping them or sending unusable arguments, they're now serialized as standard tool-use blocks, with malformed arguments falling back to an empty object. That ensures…

Second, preserving context for nested tool use. An open fix in PR 40870 addresses server tool-use blocks that run inside code execution, such as web search with dynamic filtering. Those nested calls carry a caller field that points to their parent. The message formatter had been stripping that field, making nested…

Both changes are narrowly scoped to chat models with accompanying test coverage, so release risk looks low. If you maintain Anthropic tool-calling or server tools, no action is needed, but expect fewer replay errors after these land.

What's next: watch whether PR 40870 merges, since it builds on the same replay path as…

Nearby episodes from LangChain

  1. Weekly Recap - Hardening Tool Calls and Provider Catch-Up
  2. Agent Recovery and Tool Guardrails
  3. Streaming Performance and Bedrock Compatibility
  4. Provider Guardrails and Tracing Fixes
  5. Mid-Conversation Tool Changes and Typesafe Hardening
  6. Tool Calling Hardening and Opus 5.5 Readiness
  7. Tool Reliability Fixes
  8. TypeSafe Goes Invocation-Scoped, Anthropic Caching Fix