How to set up Amazon Bedrock with Cursor

I ran out of Anthropic credits but had AWS credits sitting around. You can connect Cursor to your AWS Bedrock account to use those instead.

The problem is that the official Cursor docs (https://cursor.com/docs/settings/aws-bedrock) do not explain how to do this unless you are on Cursor Enterprise. They assume IAM role access, which you do not get on Personal or Pro plans. If you are not on Enterprise, you must use AWS access keys, and there are a few missing steps that are not documented.

This post is a straightforward guide to getting it working. The setup below is what I used successfully with Claude Sonnet 4.5 and Claude Opus 4.5. It should work for other Bedrock models as well, but I have not tested that.

What you are setting up

You will:

• Create an IAM user
• Grant it permission to invoke Anthropic models in Bedrock
• Generate access keys
• Send a test message to each model in the Bedrock playground
• Add the credentials to Cursor

Step 1: Create an IAM user

Create a new IAM user, for example:

cursor-bedrock-user

This user only needs programmatic access. It does not need console access, and you should not reuse an existing admin user.

Step 2: Attach the Bedrock permission policy

Attach the following inline policy to the IAM user:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "AllowBedrockInvokeClaude",
      "Effect": "Allow",
      "Action": [
        "bedrock:InvokeModel",
        "bedrock:InvokeModelWithResponseStream"
      ],
      "Resource": [
        "arn:aws:bedrock:*::foundation-model/anthropic.*",
        "arn:aws:bedrock:*:*:inference-profile/us.anthropic.*"
      ]
    }
  ]
}

This policy is not overly permissive, but it is sufficient. With this in place, Cursor can invoke Anthropic models using both standard and streaming inference.

Step 3: Create access keys

Generate an Access Key ID and Secret Access Key for the user.

For the description, something simple is fine:

Used by Cursor IDE to invoke Amazon Bedrock models.

Save the keys somewhere secure. You will not be able to view the secret key again.

Step 4: Send a test message in the Bedrock playground

Before configuring Cursor, open Amazon Bedrock in the AWS console and send a test message for each model you plan to use.

In my case, this was:

• Claude Sonnet 4.5
• Claude Opus 4.5

For each model:

1. Open Amazon Bedrock
2. Go to Model catalog
3. Click the model
4. Open the Playground
5. Send a simple message, such as “Hello”

You only need to do this once per model.

It may show an error the first time. That is fine. After a short delay, you should receive an email from AWS Marketplace confirming that you have subscribed to the model. Once that email arrives, the model is ready to use.

Step 5: Configure Cursor

Open Cursor and go to:

Settings → Models → Bedrock

Fill in the following fields:

• AWS Access Key ID

• AWS Secret Access Key

• AWS Region (for example us-east-1)

• Test Model ID:

  us.anthropic.claude-sonnet-4-5-20250929-v1:0

Click Validate & Save.

After this, Cursor should successfully route requests through your AWS Bedrock account.

Once set up, Cursor behaves like any other Bedrock client. Usage is billed to AWS, your credits apply normally, and you are no longer constrained by Cursor’s built-in model limits.