Auto mode unavailable when using Bedrock inference profiles

Resolved 💬 3 comments Opened Apr 30, 2026 by sinistral Closed Apr 30, 2026

Auto mode is unavailable when Claude Code is configured to use AWS Bedrock
via CLAUDE_CODE_USE_BEDROCK=1. The UI shows \"auto mode unavailable for
this model\" despite defaultMode: auto being set in settings.json and the
--enable-auto-mode flag being passed to the binary.

Configuration

  • CLAUDE_CODE_USE_BEDROCK=1
  • model: us.anthropic.claude-sonnet-4-6 (cross-region inference profile)
  • permissions.defaultMode: auto
  • Binary invoked with --enable-auto-mode
  • Claude Code version: 2.1.123

Expected behaviour

Auto mode should work with Bedrock cross-region inference profiles the same
way it works with direct Anthropic API model IDs.

Actual behaviour

\"Auto mode unavailable for this model\" is shown. Switching to the same
model via the direct Anthropic API (non-Bedrock) makes auto mode available
immediately, confirming the issue is specific to Bedrock model ID recognition.

Likely cause

The auto mode availability check appears to use a hardcoded allowlist of
Anthropic API model ID strings. Bedrock inference profile IDs
(us.anthropic.claude-sonnet-4-6, global.anthropic.claude-sonnet-4-6)
are not in that allowlist, even though they resolve to the same underlying
model.

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗