OpenAI Model Connector

Configure OpenAI model access for Harness Worker Agents with API keys, AWS, GCP, or Azure cloud provider connectors, or LiteLLM endpoints

The OpenAI Model Connector defines the model endpoint and default model for your Worker Agents. You can connect directly to OpenAI or use an OpenAI-compatible endpoint through a hosted LiteLLM proxy.

For cloud provider authentication, reference an existing AWS, GCP, or Azure connector to obtain credentials using its configured IAM role, OpenID Connect (OIDC), or other supported authentication method.

---

What you will learn in this topic

By the end of this topic, you will be able to:

• Understand which OpenAI models the connector supports.

• Configure reasoning effort levels for cost and latency tradeoffs.

• Configure authentication, the default model, and reasoning effort.

• Use cloud provider connectors with IAM roles or OIDC.

• Configure a LiteLLM endpoint and the Worker Agent backend.

• Use the Harness-managed connector if you do not have model access.

---

Before you begin

Before you configure the OpenAI Model Connector, ensure you have the following:

• Harness AI enabled: Harness AI must be active for your account. Go to Overview for more information on activating Harness AI.

• Model access: Access to the models you intend to use through OpenAI, your cloud provider, or a hosted LiteLLM server.

• Authentication credentials: For API key authentication, store your OpenAI or LiteLLM API key as a Harness text secret. For Cloud Provider, configure an AWS, GCP, or Azure connector with the roles and permissions needed for model access.

• Cloud connector access: When you use Cloud Provider, you must have permission to use the referenced cloud provider connector.

• Connector permissions: Permission to create and manage connectors in your Harness project. Go to RBAC in Harness for more information on permissions.

---

Supported models

For direct OpenAI access, the OpenAI Model Connector supports GPT-5.5 with configurable reasoning effort.

For LiteLLM, use the exact model name configured on your LiteLLM server. For example, if your deployment exposes a model as gpt-4o, enter gpt-4o as the connector model name.

---

Reasoning effort levels

The reasoning effort level controls how much reasoning the model applies before responding, trading latency and cost against depth. Set the reasoning effort level on the connector configuration form based on your task complexity requirements.

Effort level	Description
low	Fastest, lowest-cost responses for simple tasks
medium	Balanced reasoning and latency for most tasks
high	Deeper reasoning for complex tasks
xhigh	Maximum reasoning depth for the most demanding tasks

---

Connector configuration

Choose API key or cloud provider authentication, then configure your endpoint and default model.

Authentication method	Configuration
API key	Reference an OpenAI or LiteLLM API key stored as a Harness secret.
Cloud Provider	Reference an existing AWS Connector, GCP Connector, or Azure Connector to obtain credentials through its configured authentication method, including IAM roles or OIDC.

For cloud credentials, go to Cloud provider authentication. For a proxy endpoint, go to LiteLLM connector configuration to configure the URL, API key, and model name.

For direct OpenAI API key access, set the following:

• Authentication: Provide an OpenAI API key. Reference the key from a Harness secret rather than entering it in plain text.

• Model Name: Select the default model (GPT-5.5) the agent uses at runtime.

• Reasoning Effort: Select the default effort level (low, medium, high, or xhigh).

---

Cloud provider authentication

Use Cloud Provider authentication to obtain credentials through an existing Harness cloud provider connector. Configure IAM roles, OIDC federation, or other supported credentials on that connector before referencing it from the OpenAI connector.

Use the setup guide for your cloud provider:

Cloud provider selection	Credential setup
AWS Connector	Go to Add an AWS connector to configure IAM role or OIDC authentication.
GCP Connector	Go to Connect to Google Cloud Platform to configure inherited delegate credentials or OIDC authentication.
Azure Connector	Go to Add a Microsoft Azure connector to configure managed identity or OIDC authentication.

The referenced connector must have the cloud permissions required to access your selected model.

To configure the OpenAI connector:

1. In Authentication, select Cloud Provider.

2. In Select Cloud Provider, choose AWS Connector, GCP Connector, or Azure Connector.

3. In Connector, select the existing Harness connector for that provider.

4. Configure the endpoint and Model Name for your selected provider.

5. Complete the connectivity setup and connection test, then save the connector.

After you save it, select this connector as the Model Provider in your Worker Agent.

---

LiteLLM connector configuration

LiteLLM provides an OpenAI-compatible API for the models configured on your server. Worker Agents send requests to that API, and LiteLLM normalizes the requests and routes them to the configured model provider.

Before you create the connector, prepare the following:

• Hosted LiteLLM server: The server URL must be reachable from your Worker Agent execution environment.

• Configured models: The models you intend to use must be configured on the LiteLLM server.

• LiteLLM API key: The key must have access to those models. Store it as a Harness text secret in your account.

To configure the connector:

1. In Connectors, create an OpenAI connector.

2. In URL, enter your hosted LiteLLM URL with /v1 appended. For example, use https://litellm.example.com/v1 for a server hosted at https://litellm.example.com.

3. In API Key, select the Harness secret containing your LiteLLM API key.

4. In Model Name, enter the exact model name configured on the LiteLLM server, such as gpt-4o.

5. Complete the connector setup and save it.

OpenAI connector configured with a LiteLLM server URL ending in /v1, a Harness secret for the API key, and the LiteLLM model name

Local server URLs

For a LiteLLM server at http://localhost:4000, the API URL is http://localhost:4000/v1. Use this only when the server is reachable at that address from the execution environment. A Worker Agent running on Harness Cloud cannot reach a LiteLLM server on your laptop through localhost.

Use the LiteLLM connector in a Worker Agent

After you save the connector, create or edit a Worker Agent.

1. In the Visual tab, enter the agent Name.

2. Enter the agent Instructions.

3. Select any MCP connectors the agent needs.

4. For Model Provider, select the OpenAI connector you configured for LiteLLM.

5. Review and save the agent.

6. Open the YAML tab and set with.backend to openai under agent.

7. Save the updated agent.

Set the OpenAI backend explicitly

The LiteLLM setup currently requires agent.with.backend: openai. Set this value manually in the agent YAML so the agent uses the OpenAI-compatible API exposed by LiteLLM.

The following minimal agent definition shows the backend setting. Replace litellm_model with your LiteLLM model connector identifier. When you update an existing agent, keep its prompt, connectors, and inputs:

agent:
  uses: harnessAI@1.0.0
  with:
    prompt: Respond with a short greeting.
    connector: litellm_model
    backend: openai

Go to Worker Agents to review agent creation and pipeline setup.

---

Next steps

• Worker Agents: Create and configure agents that use this connector.

• Anthropic Model Connector: Configure Claude as your model provider.