Clients

Connection guide

Connect your key to Codex, Claude Desktop, or any other OpenAI-compatible app.

Quick setup

Add this line at the very beginning of config.toml.

openai_base_url = "https://keydex.shop/codex"

Configure the OpenAI-compatible provider
Save config.toml
Fully restart Codex

Auto setup

Set up Codex with one script

Your account has ready-made scripts for Windows, Linux, and macOS — they write the endpoint and your key into config.toml and auth.json. Run the script and restart Codex.

Open my account

Connection steps

The setup is identical across official Codex clients.

Open config.toml

Open Windows File Explorer, paste the path below into the address bar, and press Enter.

%USERPROFILE%\.codex\config.toml

Configure the OpenAI-compatible provider

Paste this line at the very beginning of config.toml. If openai_base_url already exists, replace its value.

openai_base_url = "https://keydex.shop/codex"

Save config.toml

Save your config.toml changes before restarting Codex.

Fully restart Codex

The settings take effect only after the application starts again.

1 Close Codex

Close the application completely.

2 Check Task Manager

If Codex is still running, end the process.

3 Launch Codex

The endpoint configuration takes effect on startup.

Sign in with your API key

When Codex starts, choose API key sign-in: “Use another sign-in method” → enter your API key → “Continue”. Done: Codex will use the configured API.

Alternative connection

Alternative endpoint

If the primary endpoint is unavailable, replace the address with the alternative one.

openai_base_url = "https://keydex.shop/v1"

Restore defaults

Return to the default connection

Remove openai_base_url from config.toml or clear the environment variable, then restart Codex.

openai_base_url = ""

Claude Desktop

Connecting Claude Desktop

Claude Desktop takes the address and key from Developer mode — no config files needed.

Gateway base URL: https://keydex.shop/codex/

Sign out of your Claude account and stay on the sign-in screen.
Turn on Developer mode: Help → Troubleshooting → Enable Developer mode.
In Developer → Configure third-party inference paste the address and key, then press Apply locally.

The setting is applied locally; no Claude account is required for it.

Key and balance

Where to get your key

Your key, remaining balance, request statistics and top-up all live in the account area.

Open account

Connection steps

Five steps for Claude for Windows.

Sign out

The third-party gateway is set up from the sign-in screen. If you are already signed in, sign out (Help → Sign out) and stay on the welcome screen — do not press “Sign in”.

Turn on Developer mode

In the Help menu choose Troubleshooting and press Enable Developer mode.

Open the gateway settings

A new Developer item appears in the menu bar. Choose Developer → Configure third-party inference.

Enter the address and key

The dialog has two fields — fill both and press Apply locally. The setting applies to this computer only.

Gateway base URL: https://keydex.shop/codex/

1 Gateway base URL

Copy the address from the block above and paste it whole, including /codex/.

2 Gateway API key

Your sk-... key — view and copy it in the account area.

3 Apply locally

The button at the bottom of the dialog. After that Claude Desktop starts sending requests through our gateway.

Check the connection

Open a new chat and ask anything: the answer comes through our gateway. Usage and balance show up in the account area.

Turn off

How to switch back

Developer → Configure third-party inference → turn the toggle off (or clear both fields) and press Apply locally. Claude goes back to account sign-in.

Not connecting

What to check

Developer mode is on; the address is pasted whole, including /codex/; the key has no stray spaces and is active in the account area.

Other IDEs and AI agents

Connect an OpenAI-compatible client

Set the Base URL and your API key in the application's provider settings.

Base URL: https://keydex.shop/codex/v1
API Key: YOUR_API_KEY

Open provider settings
Paste the Base URL and API key
Save and test the connection

Quick setup

ZCode — connecting GLM

The provider is added by hand: address, key and three GLM models.

https://keydex.shop/codex/v1

Settings → Models → Custom Provider → Add provider.
Paste the Base URL from the line below and your API Key.
Add the GLM models to Model list and save the provider.

Account

Key and balance live in the account area

The sk-… key is created and copied in the account area, where balance, per-model usage and request stats also live.

Open my account

Connection steps

Five steps; only ZCode settings change.

Open the custom providers section

Settings → Models → Custom Provider. This section is for manually added providers; official ones never show up here.

Press Add provider

ZCode creates an empty provider card — we fill it in on the next steps.

Set the Base URL

Our gateway address. It is OpenAI-compatible: no extra ports, proxies or VPN required.

https://keydex.shop/codex/v1

Paste your API Key

Copy the sk-… key from the account area — the same key works for Codex, Claude Desktop and the rest.

API Key: sk-…

Add the GLM models to Model list

List the models one per line — copy the whole list if you like.

glm-5.3
glm-5.2
glm-5.3-flash

Save the provider and pick the model you need from the model list.

If the model does not answer

What to check

The address is pasted whole, including /codex/v1; the key has no stray spaces and is active in the account area; the model is in Model list.

One key

The same key for other clients

The key works with any OpenAI-compatible client: Codex, Cursor, VS Code, OpenCode, ZCode and DeepSeek Harness — only the endpoint differs.

Quick setup

DeepSeek Harness — connecting DeepSeek

The provider is added by hand: address, key and three DeepSeek models.

https://keydex.shop/codex/v1

Settings → Models → Add provider.
Paste the Base URL from the line below and enter your API Key.
Add the DeepSeek models, save the settings and check the connection.

Account

Key and balance live in the account area

The sk-… key is created and copied in the account area, where balance, per-model usage and request stats also live.

Open my account

Connection steps

Six steps: the client settings plus the effort block in the configuration file.

Open Settings → Models

Add provider lives in the Models section, next to the list of providers you already have.

Press Add provider

Give the provider a clear name (for example keydex) — that is how it appears in the model list.

Set the Base URL

Our gateway address: OpenAI-compatible (api: openai-completions), no VPN or proxy needed.

https://keydex.shop/codex/v1

Enter your API Key

The sk-… key is in the account area; an active key is ready to use right away.

API Key: sk-…

Add the DeepSeek models to the list

Copy the whole list or type the models in one by one.

deepseek-v4-pro
deepseek-v4-flash
deepseek-v4.1-flash

Add effort (reasoningEfforts)

To get the high and max reasoning levels, open the configuration file (cordis.patch) and add reasoningEfforts to the custom provider — you can simply replace the whole block with this one:

- id: llm-pi-ai
  name: "@deepseek-ai/dsh-llm-pi-ai"
  config:
    providers:
      keydex:
        displayName: v4pro
        apiKeyEnv: KEYDEX_API_KEY
        api: openai-completions
        baseURL: https://keydex.shop/codex/v1
        models:
          - id: deepseek-v4-pro
            name: v4pro
            reasoningEfforts:
              high: high
              max: max
          - id: deepseek-v4-flash
            name: flash4
            reasoningEfforts:
              high: high
              max: max
          - id: deepseek-v4.1-flash
            name: flash4.1
            reasoningEfforts:
              high: high
              max: max

Save the file and restart the client — the effort levels then appear in the model picker.

Check the connection

A short request

Save the settings and ask the chosen model anything: if the answer arrives, the provider is set up correctly.

Choosing effort

high or max?

high is the balanced mode for everyday work; max gives the most detailed reasoning for hard tasks and spends more tokens.