Quickstart

Five steps, about 15 minutes. You need an AWS account where you can create a CloudFormation stack with an IAM role. Your requests run in that account.

1. Subscribe and claim

On the AWS Marketplace listing, choose View purchase options, pick a plan and term, and subscribe.
Choose Set up your account, sign in to the dashboard or create an account, and claim the subscription.

It shows pending until AWS confirms the purchase (usually a few minutes), then active. Carry on with the next steps meanwhile.

2. Create an API key

On API keys, create a key. It starts with acg_ and is shown once, so store it now.

3. Connect your AWS account

On AWS account, choose Launch stack in the AWS console.
Leave the parameters as they are, acknowledge the IAM role, and create the stack.
At CREATE_COMPLETE, copy RoleArn from the Outputs tab.
In the dashboard, paste it, pick a region (us-east-1, us-east-2 or us-west-2) and choose Test and save.

The role can only call and list Bedrock models. See Your AWS account.

4. Enable models, if any are listed

The dashboard lists models your account has not enabled yet. Anthropic Claude models need a one-time approval by an AWS administrator in the Bedrock console's Model catalog. Then choose Test again. Until then, requests go to the models you already have.

5. Make your first call

export ACG_API_KEY="acg_..."

curl https://api.aicloudgateway.com/v1/chat/completions \
  -H "Authorization: Bearer $ACG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "auto", "messages": [{"role": "user", "content": "What is a sports car?"}]}'

The response's model field names the Bedrock model that answered, for example amazon.nova-micro.

Error `code`	Go back to
`api_key_required`, `invalid_api_key`	Step 2
`subscription_inactive`	Step 1: wait for `active`
`bedrock_not_connected`	Step 3
`bedrock_model_not_enabled`	Step 4

Anything else: Troubleshooting.

Full guide: getting-started.md on GitHub