> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/charmbracelet/crush/llms.txt
> Use this file to discover all available pages before exploring further.

# Amazon Bedrock

> Run Anthropic Claude models through AWS Bedrock

Crush supports running Anthropic models through Amazon Bedrock, AWS's managed service for foundation models.

<Note>
  Crush currently supports Anthropic models through Bedrock with **caching disabled**.
</Note>

## Prerequisites

Before using Bedrock with Crush, you need:

1. An AWS account with Bedrock access
2. AWS credentials configured locally
3. Model access enabled in your AWS region
4. The AWS CLI installed (for `aws configure` method)

## Configuration Methods

There are two ways to configure AWS credentials for Bedrock:

### Method 1: AWS CLI Configuration

The recommended approach is to use the AWS CLI to configure your credentials:

```bash theme={null}
aws configure
```

This will prompt you for:

* AWS Access Key ID
* AWS Secret Access Key
* Default region name
* Default output format

<CodeGroup>
  ```bash AWS CLI Configuration theme={null}
  $ aws configure
  AWS Access Key ID [None]: AKIAIOSFODNN7EXAMPLE
  AWS Secret Access Key [None]: wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY
  Default region name [None]: us-east-1
  Default output format [None]: json
  ```
</CodeGroup>

### Method 2: Environment Variables

Alternatively, you can set environment variables directly:

<CodeGroup>
  ```bash Unix/Linux/macOS theme={null}
  export AWS_ACCESS_KEY_ID="AKIAIOSFODNN7EXAMPLE"
  export AWS_SECRET_ACCESS_KEY="wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY"
  export AWS_REGION="us-east-1"
  ```

  ```powershell Windows PowerShell theme={null}
  $env:AWS_ACCESS_KEY_ID="AKIAIOSFODNN7EXAMPLE"
  $env:AWS_SECRET_ACCESS_KEY="wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY"
  $env:AWS_REGION="us-east-1"
  ```
</CodeGroup>

## Required Environment Variables

Crush expects the following AWS configuration:

### AWS Region

You **must** set either `AWS_REGION` or `AWS_DEFAULT_REGION`:

```bash theme={null}
export AWS_REGION="us-east-1"
# or
export AWS_DEFAULT_REGION="us-east-1"
```

Common regions with Bedrock support:

* `us-east-1` (US East - N. Virginia)
* `us-west-2` (US West - Oregon)
* `eu-central-1` (Europe - Frankfurt)
* `ap-southeast-1` (Asia Pacific - Singapore)

<Warning>
  Bedrock is not available in all AWS regions. Check the [AWS Bedrock documentation](https://docs.aws.amazon.com/bedrock/latest/userguide/what-is-bedrock.html) for the latest regional availability.
</Warning>

## Using AWS Profiles

If you have multiple AWS profiles configured, you can specify which one to use with the `AWS_PROFILE` environment variable:

```bash theme={null}
AWS_PROFILE=myprofile crush
```

This is useful when:

* You work with multiple AWS accounts
* You need different permissions for different projects
* You want to separate personal and work credentials

### Example: Multiple Profiles

```bash theme={null}
# Use your work profile
AWS_PROFILE=work crush

# Use your personal profile
AWS_PROFILE=personal crush
```

## Alternative: Bearer Token Authentication

As an alternative to `aws configure`, you can use the `AWS_BEARER_TOKEN_BEDROCK` environment variable:

```bash theme={null}
export AWS_BEARER_TOKEN_BEDROCK="your-bearer-token"
```

This method is useful for:

* CI/CD pipelines
* Temporary credentials
* Service-to-service authentication

<Note>
  Bearer tokens are temporary credentials that expire. Ensure you refresh them as needed.
</Note>

## Enabling Bedrock in Crush

Once you have AWS configured, Crush will automatically detect your credentials and show Bedrock as an available provider. You can verify this by:

1. Running Crush: `crush`
2. Checking the list of available models
3. Looking for Bedrock/Anthropic models in the provider list

No additional configuration in `crush.json` is required for basic Bedrock usage.

## Available Models

Through Bedrock, you can access Anthropic Claude models, including:

* Claude 3.5 Sonnet
* Claude 3 Opus
* Claude 3 Sonnet
* Claude 3 Haiku

Model availability depends on your AWS region and account permissions.

## Current Limitations

<Warning>
  **Prompt caching is currently disabled** for Bedrock models in Crush.

  This means:

  * Each request will be treated as new, without cached context
  * Costs may be higher compared to direct Anthropic API usage
  * Response times may be slightly longer for repeated queries
</Warning>

## Pricing

Bedrock pricing is separate from Anthropic's direct API pricing. Key differences:

* **No prompt caching discount** (since caching is disabled in Crush)
* **Region-specific pricing** varies by AWS region
* **On-demand pricing** based on input/output tokens
* **Provisioned throughput** available for dedicated capacity

Refer to the [AWS Bedrock pricing page](https://aws.amazon.com/bedrock/pricing/) for current rates.

## Troubleshooting

### Bedrock Not Appearing

If Bedrock doesn't show up as a provider:

1. Verify AWS credentials are configured: `aws sts get-caller-identity`
2. Ensure `AWS_REGION` or `AWS_DEFAULT_REGION` is set
3. Check that your AWS account has Bedrock access enabled

### Access Denied Errors

If you see permission errors:

1. Verify your IAM user/role has Bedrock permissions
2. Request model access in the AWS Bedrock console
3. Check that your region supports Bedrock

### Region Not Supported

If your region doesn't support Bedrock:

1. Switch to a supported region (e.g., `us-east-1`)
2. Update `AWS_REGION` environment variable
3. Restart Crush

## IAM Permissions

Your AWS IAM user or role needs the following permissions:

```json theme={null}
{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "bedrock:InvokeModel",
        "bedrock:InvokeModelWithResponseStream"
      ],
      "Resource": "*"
    }
  ]
}
```

<Tip>
  Use AWS's managed policy `AmazonBedrockFullAccess` for development, but create a custom policy with minimal permissions for production.
</Tip>

## Next Steps

<CardGroup cols={2}>
  <Card title="Vertex AI" icon="google" href="/advanced/vertex-ai">
    Learn about using Google Cloud Vertex AI with Crush
  </Card>

  <Card title="Custom Providers" icon="code" href="/advanced/custom-providers">
    Configure custom API providers
  </Card>
</CardGroup>