Sign in
← AI PulseJul 24, 2026

Wire · news · Multi-source brief

Vercel AI SDK Updates Anthropic and Google Integrations, Fixes Claude Harness

Vercel AI SDK has released updates for its Anthropic and Google integrations, including support for the new Claude Opus 5 model and fixes for the Claude harness.

By Illumora Editorial

Source · Jul 24, 2026, 5:26 PM · On Illumora · Jul 24, 2026, 5:32 PM

Media from the primary source — shown here so you can stay on Illumora.

Synthesized from multiple allowlisted primaries on the same event. Lanes →

Brief drafted by Illumora’s editorial model from the linked primary source. Ops desk reviews flagged pieces. How we write →

Read the source →Vercel AI SDK Changelog — Release @ai-sdk/harness-claude-code@1.0.44 · vercel/ai
Save

Vercel AI SDK has issued several updates, including new features for its Anthropic provider, a fix for its Claude harness, and an adjustment to its Google provider. These changes were released on July 24 at approximately 17:22 and 17:25 UTC.

The @ai-sdk/anthropic@3.0.102 and @ai-sdk/anthropic@2.0.91 packages now support the Claude Opus 5 model, which offers 128k output tokens, structured output, and adaptive thinking. The Anthropic provider also gained support for fallback 'default' mode, routing safety classifier refusals to Anthropic's recommended fallback model, and mid-conversation tool changes via the toolChanges system message provider option.

Key Points

  • The @ai-sdk/anthropic@3.0.102 and @ai-sdk/anthropic@2.0.91 packages now support the Claude Opus 5 model.
  • Claude Opus 5 offers 128k output tokens, structured output, and adaptive thinking capabilities.
  • The Anthropic provider supports a 'default' fallback mode for safety classifier refusals, automatically adding the server-side-fallback-2026-07-01 beta.
  • Mid-conversation tool changes are now supported in the Anthropic provider through the toolChanges system message option, emitting tool_addition/tool_removal content blocks and the mid-conversation-tool-changes-2026-07-01 beta.
  • The @ai-sdk/harness-claude-code@1.0.44 package received a fix that removes a broken bridge channel.interrupt() layer.
  • The @ai-sdk/google@4.0.24 package now omits unsupported function call IDs.

Context

According to Vercel AI SDK changelogs, these updates enhance the functionality and stability of the SDK's integrations with Anthropic and Google models. The Anthropic updates introduce new model capabilities and improved handling of safety and tool use, while the Claude harness fix addresses a specific technical issue. The Google update refines function call handling.

Why It Matters

These updates provide developers using the Vercel AI SDK with access to advanced model features, improved error handling, and more robust tool integration. The addition of Claude Opus 5 expands the range of high-capability models available, potentially enabling more complex and nuanced AI applications.

What To Do

  • Review the Vercel AI SDK documentation for the @ai-sdk/anthropic package to understand the new Claude Opus 5 capabilities.
  • Test the 'default' fallback mode for Anthropic models to observe how safety classifier refusals are handled.
  • Experiment with the toolChanges system message provider option for mid-conversation tool adjustments in Anthropic applications.
  • Update to @ai-sdk/harness-claude-code@1.0.44 to incorporate the fix for the bridge channel.interrupt() layer.
  • Note the change in @ai-sdk/google@4.0.24 regarding the omission of unsupported function call IDs.

Keep Exploring

/atlas/claude-family /atlas/**gemini**-family /techniques/output-schema /techniques/system-user-separation /studio?pack=foundation