OpenAI

Configure OpenAI as an LLM provider in agentgateway.

Configuration

# yaml-language-server: $schema=https://agentgateway.dev/schema/config
binds:
- port: 3000
  listeners:
  - routes:
    - backends:
      - ai:
          name: openai
          provider:
            openAI:
              # Optional; overrides the model in requests
              model: gpt-3.5-turbo
      policies:
        backendAuth:
          key: "$OPENAI_API_KEY"

Multiple endpoints

Optionally, to support multiple LLM endpoints, you can set the routes field inside the ai configuration (not to be confused with the routes field under listeners).

The ai.routes field maps URL paths to one of the following route types:

• completions: Transforms to the LLM provider format and processes the request with the LLM provider. This route type supports full LLM features such as tokenization, rate limiting, transformations, and other policies like prompt guards.
• passthrough: Forwards the request to the LLM provider as-is. This route type does not support LLM features like route processing and policies. You might use this route type for non-chat endpoints such as health checks, GET requests like listing models, or custom endpoints that you want to pass through to.

The keys are URL suffix matches, like /v1/models. The wildcard character * can be used to match anything.

If no route is set, the route defaults to the completions endpoint.

In the following example, the /v1/chat/completions route is fully processed by the LLM provider. The /v1/models route and any other route (*) are passed through to the LLM provider as-is.

# yaml-language-server: $schema=https://agentgateway.dev/schema/config
binds:
- port: 3000
  listeners:
  - routes:
    - backends:
      - ai:
          name: openai
          provider:
            openAI:
              # Optional; overrides the model in requests
              model: gpt-3.5-turbo
          routes:
            /v1/chat/completions: completions
            /v1/models: passthrough
            "*": passthrough
      policies:
        backendAuth:
          key: "$OPENAI_API_KEY"

Connect to Codex

Use agentgateway as a proxy to your OpenAI provider from the Codex client.

1. Update your agentgateway configuration to include passthrough routes to the OpenAI Responses API. Leave the OpenAI provider model unset so the Codex client’s model choice is used.

   cat > config.yaml << 'EOF'
   # yaml-language-server: $schema=https://agentgateway.dev/schema/config
   binds:
   - port: 3000
     listeners:
     - protocol: HTTP
       routes:
       - backends:
         - ai:
             name: openai
             provider:
               openAI: {}
         policies:
           backendAuth:
             key: "$OPENAI_API_KEY"
           ai:
             routes:
               /v1/chat/completions: completions
               /v1/responses: passthrough
               /sse: passthrough
   EOF

2. Point Codex at agentgateway through one of the following methods.

   Environment variableCLI overrideConfig file

   Codex uses the OPENAI_BASE_URL environment variable to override the default OpenAI endpoint. Use a base URL that includes /v1 so requests go to /v1/responses and OpenAI does not return 404.

   export OPENAI_BASE_URL="http://localhost:3000/v1"
   codex

   To override the base URL for a single run, set model_provider and the provider’s name and base_url (the -c values are TOML).

   codex -c 'model_provider="proxy"' -c 'model_providers.proxy.name="OpenAI via agentgateway"' -c 'model_providers.proxy.base_url="http://localhost:3000/v1"'

   To configure the base URL permanently, add the following to your ~/.codex/config.toml. For more information, see Advanced Configuration. The name field is required for custom providers.

   [model_providers.proxy]
   name = "OpenAI via agentgateway"
   base_url = "http://localhost:3000/v1"