Support Claude 4 with AWS Bedrock

Bug Reports
1

Seemingly I always encounter this error message while trying to use the Claude 4 (Sonnet/Opus) model ID with AWS Bedrock API Key. Claude 3.7 is working but 4 just never works. Does this mean Cursor haven’t supported that?

And I am also wondering why we cannot use the common Claude model name when using AWS api key. It seems models from AWS are treated differently compared with the default Claude model?

image
image497×225 9.15 KB

2 Likes
2

Seconding this request. Strangely I am getting rate limited hard on 3.7 with bedrock, but can’t even send a single request to claude 4. Also can’t use thinking modes when using bedrock.

2 Likes
3

I am also encountering this same error when I attempt to hit Opus 4 on Bedrock. Sonnet 4, however, does work.

Version: 1.1.6

Request ID: edaab4da-0998-4139-b5e2-336a88112b16
{“error”:“ERROR_OPENAI_RATE_LIMIT_EXCEEDED”,“details”:{“title”:“Hit your Anthropic API key rate limit.”,“detail”:“You’re using your own API key, and we received a rate limit error from Anthropic. Please check your API key settings or wait a while before trying again.”,“additionalInfo”:{}},“isExpected”:true}

Separately, I see these messages when I select Sonnet 4:
Model us.anthropic.claude-sonnet-4-20250514-v1:0 doesn’t have good agent support yet. Start a new thread using Manual mode for better results.

Though, it appears that agent mode does support Sonnet 4 when I hit Anthropic directly (no bedrock).

When can we expect these issues to be fixed for Bedrock users?

@danperks

4

Hey all,

This error is likely correct, and signifies your rate limits for AWS Bedrock are too low for Cursor to be able to execute the query.

You can find more information here, including how to request higher rate limits:

5

I think I’ve resolved the rate limit issues, but I’m now seeing the following error using Sonnet 4 via Bedrock:

We encountered an issue when using your API key: messages.7.content.0.type: Expected `thinking` or `redacted_thinking`, but found `tool_use`. When `thinking` is enabled, a final `assistant` message must start with a thinking block (preceeding the lastmost set of `tool_use` and `tool_result` blocks). We recommend you include thinking blocks from previous turns. To avoid this requirement, disable `thinking`. Please consult our documentation at https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking API Error: ``` messages.7.content.0.type: Expected `thinking` or `redacted_thinking`, but found `tool_use`. When `thinking` is enabled, a final `assistant` message must start with a thinking block (preceeding the lastmost set of `tool_use` and `tool_result` blocks). We recommend you include thinking blocks from previous turns. To avoid this requirement, disable `thinking`. Please consult our documentation at https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking ```

Request ID: f91d1329-3c1a-4f45-accd-27c35abd8866

@danperks