Skip to content

01-tutorials - Add Codex CLI with OpenAI API Key Example on AgentCore Runtime #1514

@Sushant20

Description

@Sushant20

Which component is impacted by this improvement?

  • 01-AgentCore-runtime

Improvement Description
Add Codex CLI with OpenAI API Key Example on AgentCore Runtime


**Is your feature request related to a problem? Please describe.**

Currently, the agentcore-samples repository only contains examples using Claude Code with Bedrock models. There's no guidance on how to integrate third-party LLM providers (like OpenAI) with AgentCore Runtime, which limits the examples to AWS-native AI services only.

**Describe the solution you'd like**

Add a new tutorial example showing how to deploy Codex CLI (OpenAI's coding agent) on AgentCore Runtime with:
- OpenAI API key stored securely in AWS Secrets Manager
- Persistent storage using S3 Files (same pattern as Claude Code examples)
- Complete deployment lifecycle (setup, deploy, invoke, update, cleanup)
- Comprehensive documentation with architecture diagrams and troubleshooting

**Describe alternatives you've considered**

- Using environment variables for API keys (rejected - less secure)
- Using Parameter Store instead of Secrets Manager (rejected - Secrets Manager is better for credentials)
- Mounting secrets as files (rejected - more complex, less dynamic)

**Use Case**

This example benefits users who want to:
1. Use OpenAI models (GPT-4o) with AgentCore Runtime infrastructure
2. Learn how to integrate third-party LLM providers beyond Bedrock
3. Implement secure API key management for external services
4. Maintain the same persistent storage patterns across different LLM providers

**Implementation Scope**

The example will include:
- Dockerfile for Codex CLI with OpenAI configuration
- CloudFormation template for VPC and S3 Files infrastructure
- Python deployment scripts with IAM role configuration for Secrets Manager access
- Node.js HTTP server implementing AgentCore Runtime protocol
- Complete documentation following repository standards
- Error handling and troubleshooting guidance

**Similar Examples**
- `01-claude-code-with-s3-files` - Same infrastructure pattern, different LLM
- `02-claude-code-with-efs` - Similar persistent storage approach

**Additional Context**

This complements the existing coding agent examples by demonstrating:
- Multi-provider LLM support on AgentCore Runtime
- Secure external API key management patterns
- Consistent deployment patterns across different AI providers
- How to adapt the AgentCore Runtime protocol for different agent CLIs

**Labels**
- enhancement
- documentation
- tutorial
- coding-agents

Metadata

Metadata

Assignees

Labels

01-tutorials01-tutorialsdocumentationImprovements or additions to documentation

Type

No type
No fields configured for issues without a type.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions