[BUG] 400 error with custom ANTHROPIC_BASE_URL should suggest CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1

[monperrus]

Preflight Checklist

• I have searched existing issues and this hasn't been reported yet
• This is a single bug report
• I am using the latest version of Claude Code

What's Wrong?

When using Claude Code with a custom ANTHROPIC_BASE_URL (e.g. Azure AI, LiteLLM, or other gateways), Claude Code v2.1.100 sends beta headers such as advisor-tool-2026-03-01 that the third-party endpoint does not recognize, causing an immediate HTTP 400 failure with a cryptic API error.

The workaround is to set CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1, which suppresses these headers and restores functionality. However, Claude Code gives no indication that this flag exists or is relevant — the user only sees a raw 400 API error.

Related to the pattern in #22893 (same class of issue, new beta header advisor-tool-2026-03-01).

What Should Happen?

When a 400 error is received that contains "Unexpected value(s)" for the anthropic-beta header, and ANTHROPIC_BASE_URL is set to a non-Anthropic endpoint, Claude Code should output a clear, actionable message:

You appear to be using a custom API endpoint. If you are seeing beta header
errors, set CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1 to disable them.

Error Messages/Logs

API Error: 400 {"type":"error","error":{"type":"invalid_request_error",
"message":"Unexpected value(s) `advisor-tool-2026-03-01` for the
`anthropic-beta` header. Please consult our documentation at docs.claude.com
or try again without the header."},"request_id":"req_011CZudNPFEcZGR8Ace3tPTm"}

Steps to Reproduce

1. Configure Claude Code with an Azure AI endpoint:

   export ANTHROPIC_BASE_URL="https://<deployment>.services.ai.azure.com/anthropic/"
   export ANTHROPIC_AUTH_TOKEN="<api-key>"
   export ANTHROPIC_MODEL=claude-sonnet-4-6
   export ANTHROPIC_API_KEY=""
   # CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS is NOT set

2. Run claude and send any prompt.
3. Request fails with HTTP 400 — no suggestion that CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1 would fix it.

Workaround: Setting CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1 resolves the issue.

Is this a regression?

Yes — introduced in v2.1.100 (new advisor-tool-2026-03-01 beta header added without coverage by the disable flag's detection logic).

Claude Code Version

2.1.100

Platform

Linux

Additional Information

The fix could be either:

1. (Better UX) Detect Unexpected value(s) + anthropic-beta in a 400 response when ANTHROPIC_BASE_URL is custom, and surface a helpful message pointing to CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1.
2. (Stricter fix) Ensure CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1 suppresses all experimental headers including newer ones like advisor-tool-2026-03-01.

[github-actions]

Found 3 possible duplicate issues:

1. [BUG] "CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1" does not suppress anthropic-beta "advanced-tool-use-2025-11-20" #22893
2. [BUG] prompt-caching-scope-2026-01-05 beta header sent even when enablePromptCaching and useAnthropicBetaHeaders are set to false #21644
3. [BUG] 2.1.69 sends advanced-tool-use-2025-11-20 beta flag that Bedrock/LiteLLM rejects (regression from 2.1.68) #30926

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

[github-actions]

This issue has been automatically closed as a duplicate of #22893.

If this is incorrect, please re-open this issue or create a new one.

🤖 Generated with Claude Code