Quick start

This page takes you from nothing to a first answer: a key for your clients, the gateway running in Docker, and one request through each of the two main APIs. It needs Docker and an AWS account with access to a Bedrock model.

1. Choose a key for your clients

Every client sends this key to the gateway. Pick a long random value and keep it private. It is not an AWS credential.

export API_KEY="sk-$(openssl rand -hex 24)"

The gateway refuses to start without a key. In production it can read the key from SSM Parameter Store or Secrets Manager instead; see Configuration.

2. Give the gateway access to Bedrock

The simplest way is a Bedrock API key, which you create in the Bedrock console:

export AWS_BEARER_TOKEN_BEDROCK="<your Bedrock API key>"

Without it the gateway uses the standard AWS credential chain: access keys, a profile, or the role of the EC2 instance, ECS task or Lambda function it runs in. GPT-5.x and gpt-oss models need the Bedrock API key either way.

Check in the Bedrock console that the models you want are available to your account in your region.

3. Start the gateway

docker run --rm -p 8080:8080 \
  -e API_KEY -e AWS_BEARER_TOKEN_BEDROCK -e AWS_REGION=us-east-1 \
  sunerpy/bedrock-gateway-rust

The gateway listens on port 8080 and serves every route under /api/v1. In a second terminal, check that it is alive:

curl http://localhost:8080/api/v1/health

It answers OK, without a key.

4. Send a first request

Through Chat Completions:

curl -s http://localhost:8080/api/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  --json '{"model": "us.amazon.nova-2-lite-v1:0",
           "messages": [{"role": "user", "content": "Reply with exactly: BEDROCK_OK"}]}'

Through Responses:

curl -s http://localhost:8080/api/v1/responses \
  -H "Authorization: Bearer $API_KEY" \
  --json '{"model": "us.amazon.nova-2-lite-v1:0", "input": "Reply with exactly: BEDROCK_OK"}'

Use any model ID your account can call. To see them, ask the gateway:

curl -s http://localhost:8080/api/v1/models -H "Authorization: Bearer $API_KEY" | jq -r '.data[].id'

5. Point a client at it

The base URL is http://localhost:8080/api/v1 and the API key is your API_KEY. With the OpenAI Python SDK:

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8080/api/v1", api_key="sk-...your API_KEY...")
reply = client.chat.completions.create(
    model="global.anthropic.claude-sonnet-5-5",
    messages=[{"role": "user", "content": "Explain serverless in one sentence."}],
)
print(reply.choices[0].message.content)

Clients covers Codex CLI, the Node SDK and other tools.

Next steps

• Install the binary or the image for good.
• Read the configuration options.
• Choose a deployment for a server that stays up.