Background: Codex++ as a Local API Proxy
Codex is a popular AI assistant desktop application, natively designed to integrate with specific model services like OpenAI. DeepSeek AI, on the other hand, is a tech company that provides high-performance Large Language Models. To use DeepSeek models within Codex, a piece of middleware is required to bridge the two, which is the core function of the Codex++ tool.
Codex++ is an open-source project that not only offers a graphical user interface to simplify configuration but, more importantly, acts as an API Proxy on your local machine. It intercepts requests from the Codex application, converts them into the format specified by the DeepSeek API based on user configuration, sends them to DeepSeek’s servers, and then relays the response back to the Codex app. This explains why, after configuration, Codex actually connects to a local address (e.g., 127.0.0.1) instead of DeepSeek’s public API endpoint.
Prerequisites: Required Software and Environment
Before you begin the configuration, please ensure you have the following three core components ready:
- Codex Desktop Application: Download and install the version for your operating system (Windows or macOS) from its official website.
- Codex++ Management Tool: Visit its GitHub Releases page and download the appropriate installer. During installation on Windows, if you encounter a SmartScreen security prompt, select “More info” and then click “Run anyway” to proceed.
- DeepSeek API Key: Go to the DeepSeek open platform, log in, and create a new API Key. Be sure to copy and securely store this key immediately upon creation, as it will not be fully displayed again for security reasons.
After installation, launch the “Codex++ Management Tool” and check the “Overview” page to confirm that all environment checks are green, indicating a healthy base environment. It is recommended to keep the native Codex application closed during configuration.
Core Configuration: Adding the DeepSeek Provider in Codex++
All configuration is done within the “Codex++ Management Tool”; no manual editing of configuration files is necessary.
Start Provider Configuration: In the Codex++ Management Tool, navigate to the “Provider Configuration” tab and click the “+ Add Provider” button.
Fill in Provider Parameters: In the configuration dialog that appears, fill in the information precisely as follows:
- Name: Customizable, for example,
DeepSeek-API.
- Integration Mode: Must be set to “API-Only” mode.
- BaseUrl: Enter DeepSeek’s official API endpoint
https://api.deepseek.com/v1.
- Key: Paste the DeepSeek API Key you obtained during the prerequisite stage.
- Upstream Protocol: Must be switched to “Chat Completions”.
- Configure Models: Can be left blank for now, or you can pre-fill it with a target model, such as
deepseek-v4-pro.
Fetch and Confirm Model List: Scroll down to the “Model List” section and click the “Fetch from Upstream” button. The tool will use your provided API Key to automatically request and populate the list of available DeepSeek models. Ensure that the model you wish to use (e.g., deepseek-v4-pro) appears in the list.
Save Configuration: After confirming that all information is correct, click the “Save” button at the top of the page. You will be returned to the provider list, where you will see the newly created DeepSeek provider.
Activation and Verification
Once configured, you need to activate the provider and restart the service for the settings to take effect.
Enable Provider: In the provider list, find the DeepSeek entry you created and click the “Use” button on the right to activate it as the current API source.
Restart Service: Return to the “Overview” page and click the “Restart Codex++” button in the top-right corner. This action will apply all changes and automatically launch the Codex desktop application. Note: You must start the application using this button; opening the Codex shortcut on your desktop directly will not load the new configuration. If the application does not launch automatically after clicking, try restarting your computer and then performing this step again.
Verify Success: When the Codex application launches, you should be able to see and select models from DeepSeek (e.g., deepseek-v4-pro) in the model selection dropdown menu in the main chat interface. At this point, the integration is successful.
Technical Limitations and Considerations
Although this method successfully connects Codex and DeepSeek, users should be aware of the technical capability differences between them.
- Model Capabilities: DeepSeek’s models are primarily powerful language models, excelling in text generation, comprehension, and coding. However, they are not native multimodal models. If you use certain advanced features in Codex that require screen comprehension or image analysis (e.g., a “Computer Control” plugin), it may lead to task failure or an ineffective loop of attempts due to a mismatch in model capabilities.
- Cost Control: When a model repeatedly attempts a task it is not equipped to handle, it will continuously make API calls, which can quickly consume your API credits. Therefore, before using a DeepSeek model for complex tasks, assess whether the task type aligns with the model’s capabilities, and monitor your API usage during the process to avoid unnecessary expenses.
