Configure ChatGPT Work/Desktop, Codex CLI and Codex IDE extensions to use Amazon Bedrock runtime with AWS profile
Overview
The ChatGPT desktop app is a conversational AI integrated application from OpenAI for macOS, Windows, and Linux. Codex CLI is an AI terminal-native coding agent created by OpenAI that reads, edits, and execute code. Both ChatGPT Work and Codex support OpenAI frontier models on Amazon Bedrock Runtime and Mantle endpoints. This article covers how to configure them to use Bedrock Runtime
This article covers ChatGPT/Codex. Similar articles are available for Claude Code CLI and Claude Desktop
Requirements
You will need the following
- Amazon Bedrock access to OpenAI models
- AWS CLI installed and configured
- IAM permission access to Bedrock
Bedrock model access
Request Bedrock model access by selecting and interacting with a ChatGPT frontier model from the model catalog in the Amazon Bedrock console playground in your desired AWS Region.
Create a support case if you are unable to use any models.
IAM permission / IAM Identity Center permission set
Your AWS profile must have the following permissions
{
"Version": "2012-10-17",
"Statement": [
{
"Action": [
"bedrock:InvokeModel",
"bedrock:InvokeModelWithResponseStream",
"bedrock:ListInferenceProfiles",
"bedrock:GetInferenceProfiles"
],
"Resource": "*",
"Effect": "Allow"
},
{
"Condition": {
"StringEquals": {
"aws:CalledViaLast": "bedrock.amazonaws.com"
}
},
"Action": [
"aws-marketplace:Subscribe",
"aws-marketplace:ViewSubscriptions"
],
"Resource": "*",
"Effect": "Allow"
}
]
}
If you are using IAM Identity Center (IdC), create a permission set with the above as inline policy. Assign your user or group access to your AWS account with the created permission set.
Install AWS CLI
Installation files
To install, you can download latest binaries/installer
Do refer to AWS CLI documentation for guidance and other installation options.
To verify
aws --version
Output should be similar to below
aws-cli/2.35.20 Python/3.14.6 Windows/11 exe/AMD64
Configure AWS CLI profile
You need to configure AWS credentials on your local machine. The server uses these credentials to authenticate your requests.
Do secure your AWS credentials as credential theft can lead to LLMJacking attacks
Option 1: AWS Management Console credentials
If you have username and password that allows you to sign in to the console.
aws login
Option 2: AWS SSO (IAM Identity Center)
If you login using organization instance of IAM Identity Center
aws configure sso
Follow the prompts to set up as per your Identity Center configuration
SSO session name (Recommended): BedrockInference
SSO start URL [None]: https://d-12345678.awsapps.com/start
SSO region [None]: ap-southeast-1
SSO registration scopes [sso:account:access]:
Attempting to open your default browser. If the browser does not open, open the following URL.
If you are unable to open the URL on this device, run this command again with the '--use-device-code' option.
Option 3: IAM User
If you have a IAM user
aws configure
Enter your Access Key ID, Secret Access Key, default region, and output format.
Option 4: EC2 Instance
On EC2 instances with no AWS profiles configured, AWS CLI automatically and securely retrieves the credentials from the instance metadata without additional configuration.
If you wish to configure a profile, for example EC2, that uses EC2 instance metadata
aws configure set --profile EC2 credential_source Ec2InstanceMetadata
aws configure set --profile EC2 region ap-southeast-1
Do replace ap-southeast-1 with the correct Region Code
Verify AWS profile
Do note the profile name. If you do not specify one, the profile name is usually default. You can retrieve profile name with the following command
List Profiles
aws configure list-profiles
Output could be similar to below
BedrockInference
/aws/config file
Or you can view contents of ~/.aws/config file. Example below is a AWS SSO configured profile with name BedrockInference
[profile BedrockInference]
sso_session = BedrockInference
sso_account_id = 123456789012
sso_role_name = BedrockInference
region = ap-southeast-1
[sso-session BedrockInference]
sso_start_url = https://d-12345678.awsapps.com/start
sso_region = ap-southeast-1
sso_registration_scopes = sso:account:access
Verify Bedrock access
Run the following command (using correct AWS profile name) to list OpenAI models
aws bedrock list-inference-profiles --query "inferenceProfileSummaries[?status=='ACTIVE' && contains(inferenceProfileId, 'openai.gpt')].[inferenceProfileId]" --output text --profile BedrockInference

Refer to documentation for supported models
IAM Identity Center login
For IdC, you may have to login first
aws sso login --profile BedrockInference
Install ChatGPT and Codex
ChatGPT Desktop App
Desktop app can be installed from https://learn.chatgpt.com/docs/app
Close ChatGTP app when prompted to sign-in

Important: go to task notification bar. Click ChatGPT and select Exit

Codex CLI
Codex CLI can be installed from https://learn.chatgpt.com/docs/codex/cli
PS C:\ > powershell -ExecutionPolicy ByPass -c "irm https://chatgpt.com/codex/install.ps1 | iex"
==> Installing Codex CLI
==> Detected platform: Windows (x64)
==> Resolved version: 0.157.1
==> Downloading Codex CLI
==> PATH updated for future PowerShell sessions.
==> Current PowerShell session: codex
==> Future PowerShell windows: open a new PowerShell window and run: codex
Codex CLI 0.157.1 installed successfully.
Start Codex now? [y/N]: y
Close terminal when prompted to sign in

Codex IDE extension
IDE extensions can be installed from https://learn.chatgpt.com/docs/codex/ide
Configure ~/.codex/config.toml
Configure Bedrock Runtime
Use an editor to edit ~/.codex/config.toml
Add the following text with appropriate values for model, profile and region. (Refer to ChatGPT documentation for supported models)
model_provider = "amazon-bedrock-runtime"
model = "global.openai.gpt-6-luna"
model_reasoning_effort = "medium"
[model_providers.amazon-bedrock-runtime.aws]
profile = "BedrockInference"
region = "ap-southeast-1"

Save the file
Start ChatGPT/Codex
ChatGPT
You may be prompted with the welcome wizard. Proceed to main screen

Codex CLI

VS Code Codex IDE Extension

More information
Refer to official documentation Use ChatGPT Work and Codex with Amazon Bedrock for up-to-date guidance including other configuration methods
Feature availability
Refer to documentation for feature availability
Agent Toolkit for AWS
To give ChatGPT and Codex the tools, knowledge, and guardrails to build production-ready infrastructure on AWS, you can install Agent Toolkit for AWS. Refer to this article for installation guidance