ToolsSeptember 3, 2026via AWS Machine Learning Blog
Set up OpenAI ChatGPT Codex with LiteLLM on Amazon ECS and Amazon Bedrock
Why it matters
A vendor how-to with real engineering detail on orchestrating model access across OpenAI and Bedrock. Useful for teams managing API sprawl and cost control, but narrow in scope — deployment pattern rather than a platform shift.
Key signals
- LiteLLM gateway deployed on Amazon ECS with AWS Fargate
- Routes requests through Responses API with scoped identities and rate limits
- Compares direct IAM Identity Center access vs. managed Portkey deployment
- Connects OpenAI models via Amazon Bedrock
- Includes budgets, rate limits, and telemetry configuration
- LiteLLM gateway deployed on ECS with Fargate
- OpenAI model access through Amazon Bedrock
- Codex routing through Responses API
- Identity scoping and budget controls
- Rate limiting and telemetry
- Comparison with IAM Identity Center and Portkey managed deployment
The hook
AWS shows how to run a multi-model gateway on ECS — scoped identities, budgets, rate limits, all managed.
Deploy a customer-operated LiteLLM gateway on Amazon ECS with AWS Fargate, connect it to an OpenAI model on Amazon Bedrock, and configure Codex to route requests through the gateway's Responses API with scoped identities, budgets, rate limits, and telemetry. We also compare direct IAM Identity Cente…