claude-bedrock-client

Invoke Anthropic Claude models through AWS Bedrock with a simple, clean API. Built on top of @anthropic-ai/bedrock-sdk.

Installation

npm install claude-bedrock-client

AWS Credential Setup

This package reads credentials from environment variables or your ~/.aws/credentials file — no credentials are ever hardcoded.

Option 1 — Environment variables

export AWS_ACCESS_KEY_ID=your-access-key
export AWS_SECRET_ACCESS_KEY=your-secret-key
export AWS_REGION=us-east-1

Or in a .env file (add .env to your .gitignore):

AWS_ACCESS_KEY_ID=your-access-key
AWS_SECRET_ACCESS_KEY=your-secret-key
AWS_REGION=us-east-1

Option 2 — AWS credentials file

Configure credentials with the AWS CLI:

aws configure

This writes to ~/.aws/credentials and is picked up automatically.

IAM permissions required

Your IAM user or role needs:

{
  "Effect": "Allow",
  "Action": ["bedrock:InvokeModel", "bedrock:InvokeModelWithResponseStream"],
  "Resource": "arn:aws:bedrock:*::foundation-model/anthropic.*"
}

Make sure Claude model access is enabled in the AWS Bedrock console for your region.

Usage

Basic chat

const { invokeClaude } = require('claude-bedrock-client');

const response = await invokeClaude([
  { role: 'user', content: 'What is the capital of France?' }
]);

for (const block of response.content) {
  if (block.type === 'text') console.log(block.text);
}
// → Paris is the capital of France.

Multi-turn conversation

const { invokeClaude } = require('claude-bedrock-client');

const response = await invokeClaude(
  [
    { role: 'user',      content: 'My name is Ankit.' },
    { role: 'assistant', content: 'Hello Ankit! Nice to meet you.' },
    { role: 'user',      content: "What's my name?" },
  ],
  {
    system: 'You are a helpful assistant.',
    maxTokens: 256,
  }
);

Streaming

const { streamClaude } = require('claude-bedrock-client');

const finalMessage = await streamClaude(
  [{ role: 'user', content: 'Write a haiku about the ocean.' }],
  {
    onChunk: (text) => process.stdout.write(text),
  }
);

console.log(`\nTokens — in: ${finalMessage.usage.input_tokens}, out: ${finalMessage.usage.output_tokens}`);

Custom model / region

const { invokeClaude } = require('claude-bedrock-client');

const response = await invokeClaude(
  [{ role: 'user', content: 'Hello' }],
  {
    model: 'us.anthropic.claude-haiku-4-5',
    maxTokens: 512,
    region: 'us-west-2',
  }
);

Low-level client access

const { createClient } = require('claude-bedrock-client');

const client = createClient({ region: 'eu-central-1' });
// client is a fully configured AnthropicBedrock instance
const response = await client.messages.create({ ... });

API

invokeClaude(messages, options?)

| Parameter | Type | Default | Description | |-----------|------|---------|-------------| | messages | Array<{role, content}> | required | Conversation turns | | options.model | string | us.anthropic.claude-sonnet-4-6 | Bedrock model ID | | options.maxTokens | number | 1024 | Max output tokens | | options.system | string | — | System prompt | | options.region | string | AWS_REGION env / us-east-1 | AWS region | | options.accessKey | string | AWS_ACCESS_KEY_ID env | AWS access key | | options.secretKey | string | AWS_SECRET_ACCESS_KEY env | AWS secret key |

Returns: Promise<BedrockResponse>

streamClaude(messages, options?)

Same options as invokeClaude, plus:

| Parameter | Type | Default | Description | |-----------|------|---------|-------------| | options.maxTokens | number | 512 | Max output tokens | | options.onChunk | Function | — | Called with each text chunk |

Returns: Promise<FinalMessage> (includes usage stats)

createClient(options?)

Returns a raw AnthropicBedrock client for advanced usage.

Constants

const { DEFAULT_MODEL, DEFAULT_REGION } = require('claude-bedrock-client');
// DEFAULT_MODEL  → 'us.anthropic.claude-sonnet-4-6'
// DEFAULT_REGION → 'us-east-1'

Running the example

# Copy and fill in your credentials
cp .env.example .env  # edit .env with your AWS keys

node example/index.js

License

MIT