Configure the ChatGPT desktop app with Amazon Bedrock in AWS GovCloud (US)
The guide shows how to configure the ChatGPT desktop app to route OpenAI model inference through Amazon Bedrock Mantle in the AWS GovCloud (US‑West) region (us‑gov‑west‑1), enabling secure, token‑based billing, zero‑data‑retention processing, and IAM‑controlled access. It details a three‑step setup—authentication, configuration files, and launch/verification—while restricting network access to the approved Bedrock endpoint. Engineers can thus deploy the ChatGPT desktop app in highly regulated environments without exposing data to external storage or unmanaged endpoints.
⚡ Key Takeaways
- The deployment routes model requests to the Amazon Bedrock Mantle endpoint in **us‑gov‑west‑1**.
- Security is enforced via **AWS Identity and Access Management (IAM)** and a managed device policy that limits network connections to the Bedrock endpoint.
- **Zero Data Retention (ZDR)** ensures prompts and responses are not written to durable storage by AWS or shared with model providers.
- Integration points include the **ChatGPT desktop app**, **Codex CLI**, and IDE plugins for **Visual Studio Code**, **JetBrains**, and **Xcode**.
- The solution requires a **named AWS profile** and uses the **Responses API** on Amazon Bedrock for inference.
This configuration lets agencies meet strict data‑governance and residency requirements while providing developers with AI‑assisted coding and multi‑step task capabilities, all under existing GovCloud IAM and network controls.
✅ Practical Steps
- Follow the three‑step walkthrough: set up authentication with a named AWS profile, create the required ChatGPT desktop app configuration files pointing to the Bedrock Mantle endpoint in us‑gov‑west‑1, then launch and verify the app.
- Apply the managed device policy that restricts network access to only the approved Amazon Bedrock endpoint.
- Use the Codex CLI or IDE integrations (VS Code, JetBrains, Xcode) to invoke the configured ChatGPT desktop app for coding or multi‑step tasks.
Want the full story? Read the original article.
Read on AWS ML Blog ↗