PeakkAI Integration with Codex Desktop Tutorial

This tutorial applies to:
OpenAI Codex Desktop
PeakkAI API
macOS
Windows
OpenAI Compatible API

I. What is Codex Desktop

Codex Desktop is a desktop AI programming assistant launched by OpenAI.

Compared to traditional chat windows, Codex Desktop supports:

Code repository analysis

AI Agents

Automatic project modification

Shell execution

Git operations

Multi-task parallelism

Local tool invocation

Developers can directly use their own API service providers without being forced to use the official OpenAI API.

II. Does PeakkAI Support Codex Desktop?

Yes.

PeakkAI provides an OpenAI Compatible/Responses API.

Therefore, it can be directly integrated into Codex Desktop as a:

Custom Provider

OpenAI Compatible Endpoint

Integration can be completed simply by configuring the:

API Key

Base URL

Model

III. Prerequisites

Prepare the following items:

Item	Description
Codex Desktop	Installed
PeakkAI API Key	sk-xxxx
PeakkAI API URL	https://api.peakk.ai/v1
Model Name	gpt-5.2

IV. Download and Install Codex Desktop

Go to the official OpenAI download page to install the latest version of Codex Desktop:
https://openai.com/zh-Hans-CN/codex/get-started/
Launch the application after the installation is complete.

V. Get PeakkAI API Information

Go to the PeakkAI Open Platform to apply for your secret key (sk).

peakk Open Platform

Obtain the following information:

Configuration Item	Example
Base URL	https://api.peakk.ai
API Key	sk-xxxx
Model	Claude Opus 4.6 / GPT 5.2 ...

VI. Configuration via Environment Variables (The OpenAI Recommended Integration Method)

Codex CLI essentially uses the OpenAI SDK, so you only need to configure environment variables.
However, this method is only applicable to OpenAI-related models.
Supported examples: GPT-5.2, GPT-5.5
Unsupported examples: Claude Haiku 4.5, Gemini 2.5 Pro (These are not supported via this method)

1. Mac / Linux Configuration

Open your terminal:

Add the following lines:

Save the file and run:

2. Windows Configuration (Command Line Method)

Open either:

CMD

PowerShell

Windows Terminal

Then execute the following commands.

CMD Configuration Method

Set API Key

setx OPENAI_API_KEY "sk-xxxx"

Set Base URL

setx OPENAI_BASE_URL "https://api.peakk.ai/v1"

Set Model

setx OPENAI_MODEL "gpt-5.2"

After Configuration

Close the current terminal window.

Reopen CMD or PowerShell.

Verify the configuration:

echo %OPENAI_API_KEY%
echo %OPENAI_BASE_URL%
echo %OPENAI_MODEL%

VII. Verify Configuration

Open Codex Desktop.
Confirm whether the API has been correctly configured and test if it works properly, as shown below:

Alternatively, type a prompt like:

Help me generate a Spring Boot sample project.

If it returns the results successfully, it indicates that:

The API Key is correct.

The Base URL is correct.

The Model is correct.

PeakkAI has been successfully integrated.

VIII. Troubleshooting / FAQ

Unauthorized

Check if the:

API Key is correct

Provider Connection Failed

Check whether the URL:

https://api.peakk.ai/v1

is entered correctly.

Model Not Found

Check the model name.

Make sure you have entered:

gpt-5.2

or any other model supported by the PeakkAI platform.

IX. Summary

By utilizing the Custom Provider feature in Codex Desktop, you can directly integrate the PeakkAI OpenAI Compatible API.

Once configuration is complete, you can use models like:

GPT-5.2

GPT-5.5

in Codex Desktop, achieving a unified API and a multi-model AI programming experience.

PeakkAI API Integration with Codex Desktop

PeakkAI Integration with Codex Desktop Tutorial#

I. What is Codex Desktop#

II. Does PeakkAI Support Codex Desktop?#

III. Prerequisites#

IV. Download and Install Codex Desktop#

V. Get PeakkAI API Information#

VI. Configuration via Environment Variables (The OpenAI Recommended Integration Method)#

1. Mac / Linux Configuration#

2. Windows Configuration (Command Line Method)#

CMD Configuration Method#

Set API Key#

Set Base URL#

Set Model#

After Configuration#

VII. Verify Configuration#

VIII. Troubleshooting / FAQ#

Unauthorized#

Provider Connection Failed#

Model Not Found#

IX. Summary#