TestingOps.ai

AI model providers · Model provider

OpenAI and TestingOps: run test agents on your OpenAI API key

Give TestingOps your OpenAI API key and the agents on your projects run OpenAI models through the Codex adapter, billed to your OpenAI account.

  • Model provider
  • Available
Request a demo OpenAI API quickstart
Connection
Model provider
Writes back
Nothing, only model requests
Last checked
7 Nov 2026

In short

OpenAI and TestingOps in short

The OpenAI connection lets your organisation run TestingOps agents on its own OpenAI API key, which the customer admin saves under Integrations → AI models and TestingOps attaches, encrypted, to all of your projects.

With OpenAI first in your provider list, your agents move to the Codex adapter and use your key and the model you choose.

A Codex run on your projects needs your OpenAI key or another OpenAI-compatible key of yours, such as Azure OpenAI, a package AI allowance or permission to use the platform key; without any of these it stops before a model request is sent.

What it does

What it does for your testing

  1. 1

    Agents on your OpenAI account

    Input
    Your OpenAI API key, first under Your providers
    TestingOps
    Switches your organisation's agents to the Codex adapter and attaches the key as OPENAI_API_KEY to all of your projects
    Output
    Agent runs whose OpenAI usage is billed to your account
  2. 2

    A model from OpenAI's own list

    Input
    Your saved key
    TestingOps
    Asks OpenAI which models the key can use and offers them in the Model field
    Output
    A model name that matches what your account can call
  3. 3

    Rejections shown before the run

    Input
    A key OpenAI refuses, for example after it was revoked
    TestingOps
    Checks the key against OpenAI before a run on it and keeps the result for five minutes
    Output
    The run does not start, and the task shows OpenAI's own message instead of a generic authentication error

Step by step

How it works

  1. Save the keyThe customer admin picks OpenAI under Add a model provider in the LLM card (Integrations → AI models), pastes the key, optionally chooses a model and saves.
  2. Attached to every projectTestingOps stores the key encrypted and attaches it to every project of your organisation as OPENAI_API_KEY. New projects get it too.
  3. Agents follow the first providerWhen OpenAI is marked Default · drives agents, your agents switch to the Codex adapter at their next run. A run already in progress keeps its adapter.
  4. Checked before each runBefore a run on your key, TestingOps asks OpenAI for its model list with the key. If OpenAI refuses for a reason other than an empty balance, the run stops before it starts and shows OpenAI's message.

Connecting ChatGPT to TestingOps, so you can use TestingOps from a ChatGPT chat, is a separate connection guide. The key on this page comes from the OpenAI API platform.

Connect

Setup

  1. Create an API key on the OpenAI API platform.
  2. In TestingOps, as customer admin, open Integrations → AI models and pick OpenAI under Add a model provider.
  3. Paste the key into API key, choose a Model from the suggestions if agents should use one model, and select Save.
  4. Open the entry and select Test connection to send one short request with the key and model.
  5. Check that OpenAI shows Default · drives agents under Your providers; otherwise move it up and select Save order.

Documentation: OpenAI API quickstart

Where you connect it in TestingOps: Integrations → AI models → Your providers → OpenAI

Use cases

Use cases

Run test agents under your OpenAI organisation

Trigger
Your company allows AI usage only on its own OpenAI organisation
Uses
Your OpenAI API key, first under Your providers
Agent activity
Agents writing and reviewing cases on your projects run through Codex with your key
Result
Agent usage billed to your OpenAI organisation
Condition
Some TestingOps AI features outside agent runs use the platform's models

Pin the agents to one OpenAI model

Trigger
You want every agent on your projects on the same model
Uses
The Model field, filled from OpenAI's model list for your key
Agent activity
Your agents move to the chosen model at their next run
Result
Comparable results across agents and runs

Move from Azure OpenAI to OpenAI, or back

Trigger
Your company changes which OpenAI service it pays for
Uses
Your provider list
Agent activity
After you remove the old provider and add the new one, the Codex agents use the new key from their next run
Result
The same adapter and agents on the other service
Condition
Both write the same key variable on your projects, so keep only one of them

Capabilities

Capabilities and limits

What is read

  • OpenAI's model list for your key, for the Model suggestions and the check before a run
  • Model responses while agents work on your projects

What is produced

  • Agent runs on your projects that use OpenAI models through Codex
  • A Reachable or Unreachable mark after Test connection

Where it can write

Nothing in your OpenAI account; TestingOps only sends model requests with your key.

How it connects

OpenAI's API with your key, through the Codex adapter. A linked RunningOps organisation receives a copy of the key and the provider order.

Account and permissions

  • An OpenAI API key from the OpenAI API platform
  • Customer admin rights in TestingOps, or a company admin of the TestingOps workspace managing keys for your organisation

Limits

  • OpenAI and the OpenAI-compatible providers (Azure OpenAI, Qwen, Kimi, DeepSeek, Ollama, self-hosted) write the same OPENAI_API_KEY variable on your projects: keep only one of them per organisation
  • One key per organisation for all of its projects
  • Not every TestingOps AI feature uses your key: some, such as generating cases from a prompt, run on the platform's models

Example

Example: how a team uses it

Illustrative scenario, not a customer story.

A fintech team may send test data only to the company's own OpenAI organisation. Its customer admin saves the organisation's API key, picks a model from the list OpenAI returns for that key and leaves OpenAI first under Your providers.

The team's agents now run through Codex on that key. When the company later moves to Azure OpenAI, the admin removes the OpenAI key before adding the Azure endpoint, because both use the same key variable on the projects.

Frequently asked questions

Is this the ChatGPT integration?

No. This page is about the models TestingOps agents run on. Using TestingOps from ChatGPT, where ChatGPT calls TestingOps' MCP server, has a connection guide of its own.

Which adapter do OpenAI agents use?

Codex. When OpenAI is your first provider, your agents switch to the Codex adapter at their next run and read your key from the project.

Can we use OpenAI and Azure OpenAI at the same time?

Not for one organisation. Both write the same OPENAI_API_KEY variable on your projects, so keep the one you pay for and remove the other.

What if the key is missing or rejected?

Without an OpenAI or OpenAI-compatible key of yours, a Codex run starts only with a package AI allowance or permission to use the platform key; otherwise it stops before any request, with a message that names the missing key. If OpenAI rejects your key, the run does not start and the task shows OpenAI's message.

Is all of our data processed under our OpenAI account?

Agent runs on your projects are, while OpenAI is your first provider. Some TestingOps AI features use the platform's models, so we do not promise more than that.

Next step

Request a demo to see agents on your projects run through Codex on your own OpenAI key.

Request a demoMore ai models integrations