RouterPlex Models for VS Code

Use the RouterPlex model catalog directly in VS Code Chat and configure OpenAI's Codex extension with the same RouterPlex API key.

RouterPlex provides one OpenAI-compatible API for GPT, Claude, Gemini, Kimi, DeepSeek, Qwen, and other models with prepaid billing and per-key spend limits.

Requirements

• VS Code 1.105 or newer
• A RouterPlex account
• A funded, dedicated RouterPlex API key
• The official Codex extension if you want the Codex integration

Install

Install the released VSIX

1. Download routerplex-models-0.1.3.vsix from the latest GitHub release.
2. Open VS Code.
3. Open the Command Palette with Ctrl+Shift+P or Cmd+Shift+P.
4. Run Extensions: Install from VSIX....
5. Select the downloaded VSIX and reload VS Code when prompted.

You can also install it from a terminal when the code command is available:

code --install-extension routerplex-models-0.1.3.vsix

Build from source

git clone https://github.com/MaridWSH/routerplex-vscode.git
cd routerplex-vscode
npm install
npm test
npm run check
npm run package

Install the generated routerplex-models-0.1.3.vsix through Extensions: Install from VSIX....

First-time setup

1. Select the RouterPlex icon in the Activity Bar.
2. Expand Actions and select Run Setup.
3. Paste a dedicated RouterPlex API key.
4. Choose the default RouterPlex model for Codex.
5. Reload VS Code before opening a new Codex chat.

The setup command performs a connection test before saving the key. It then:

• registers the live RouterPlex chat-model catalog in VS Code;
• configures RouterPlex as a Codex custom model provider;
• exports ROUTERPLEX_API_KEY for Codex and new integrated terminals;
• creates a backup before changing Codex's config.toml;
• keeps the API key out of config.toml.

RouterPlex control panel

Select the RouterPlex icon in the Activity Bar to open the branded control panel. The compact panel uses RouterPlex colors while keeping every action inside VS Code.

The panel provides:

• API-key and Codex configuration status;
• the current Codex model;
• a searchable live RouterPlex model catalog;
• one-click Codex model switching;
• API-key replacement and connection testing;
• model refresh and extension update checks;
• Codex configuration, RouterPlex settings, and configuration removal.

Expand Codex Models and select any model to configure it for Codex. The panel refreshes after configuration changes and when the live catalog changes.

Use RouterPlex in VS Code Chat

1. Open VS Code Chat.
2. Open the model picker.
3. Select a model whose provider is RouterPlex.
4. Send a prompt normally.

The extension uses RouterPlex's OpenAI-compatible Chat Completions endpoint for VS Code Chat. Streaming responses and tool calls are supported. The model list comes from RouterPlex's live public catalog. It refreshes automatically every five minutes and whenever VS Code regains focus, so newly added models appear without reinstalling the extension. You can also run RouterPlex: Refresh Models for an immediate refresh.

Use RouterPlex in Codex

Run RouterPlex: Configure Codex to select or change the active model. The extension writes a managed RouterPlex provider to the user-level Codex configuration and uses the Responses API expected by Codex.

The generated provider is equivalent to:

model_provider = "routerplex"
model = "gpt-5.6-sol"

[model_providers.routerplex]
name = "RouterPlex"
base_url = "https://api.routerplex.com/v1"
wire_api = "responses"
env_key = "ROUTERPLEX_API_KEY"
env_key_instructions = "Export ROUTERPLEX_API_KEY before starting VS Code or Codex."

The extension exports ROUTERPLEX_API_KEY in the detected shell profile on macOS and Linux (.bashrc, .zshrc, fish config, or .profile). On Windows it sets a persistent user environment variable. It also injects the variable into new VS Code integrated terminals and the current extension host. Reload VS Code and start a new Codex chat after configuration.

The Codex CLI and official VS Code extension share the same provider configuration and environment variable.

Automatic extension updates

Starting with version 0.1.2, RouterPlex Models checks the latest stable GitHub Release every six hours. When a newer matching VSIX is available, the extension downloads and installs it through VS Code, then asks you to reload the window.

You can run RouterPlex: Check for Updates at any time. Set routerplex.autoUpdate to false to disable background installation. Releases are only accepted from the official MaridWSH/routerplex-vscode repository and must contain a version-matched routerplex-models-<version>.vsix asset.

Users upgrading from 0.1.1 need to install 0.1.2 once. Later releases can then update automatically.

Security

• VS Code Chat uses VS Code's encrypted SecretStorage.
• Codex reads ROUTERPLEX_API_KEY from the process environment.
• macOS and Linux persist the export in the detected shell profile.
• Windows persists the variable in the current user's environment.
• The API key is never written into Codex's config.toml.
• Existing Codex configuration is backed up before every managed change.
• RouterPlex: Remove Configuration removes the SecretStorage key, managed environment export, and Codex settings while restoring the previous model.

Use a separate RouterPlex key for editor and agent usage. Give it a hard budget and only the model access required for your workflow.

Commands

Command	Purpose
RouterPlex: Open Control Panel	Open the RouterPlex Activity Bar panel
RouterPlex: Set Up VS Code and Codex	Complete first-time setup
RouterPlex: Manage Connection	Open the RouterPlex management menu
RouterPlex: Configure API Key	Add or replace the API key
RouterPlex: Configure Codex	Choose the Codex model and update config
RouterPlex: Test Connection	Validate the stored key
RouterPlex: Refresh Models	Reload the public model catalog
RouterPlex: Check for Updates	Check, download, and install the latest release
RouterPlex: Open Codex Configuration	Open the active config.toml
RouterPlex: Open Settings	Open extension settings
RouterPlex: Remove Configuration	Remove credentials and managed settings

Settings

Setting	Default	Purpose
routerplex.apiBaseUrl	https://api.routerplex.com/v1	Inference and Codex API base URL
routerplex.catalogUrl	https://routerplex.com/api/models	Public live model catalog
routerplex.codexHome	Empty	Override CODEX_HOME or ~/.codex
routerplex.shellProfile	Empty	Override the shell profile used for ROUTERPLEX_API_KEY
routerplex.modelRefreshIntervalMinutes	5	Automatic model catalog refresh interval
routerplex.autoUpdate	true	Automatically install stable GitHub release updates
routerplex.updateCheckIntervalHours	6	Automatic extension update check interval

Troubleshooting

The API key is rejected

Create a new key in the RouterPlex dashboard and make sure the account has available credit. Run RouterPlex: Configure API Key, then RouterPlex: Test Connection.

RouterPlex models are missing from VS Code Chat

Confirm that VS Code is version 1.105 or newer, then run RouterPlex: Refresh Models and reload the window.

Codex still uses the previous provider

Run RouterPlex: Configure Codex, fully reload VS Code, and start a new Codex chat. Use RouterPlex: Open Codex Configuration to inspect the active file. In a terminal, verify the export with printenv ROUTERPLEX_API_KEY without copying the value into logs or support messages.

Remote, WSL, or container development

Run the setup command in the VS Code window connected to that environment. The extension configures the Codex home and shell profile belonging to the extension host where it is running. This also avoids relying on VS Code's global-storage URI being a local file: URI.

Development

npm install
npm test
npm run check
npm audit --audit-level=high
npm run package

The source is bundled with esbuild. Unit tests cover TOML merging and removal, environment export management, SSE parsing, tool-call translation, and catalog handling.

Links

• RouterPlex
• API documentation
• Model catalog
• API keys and budgets
• Report an issue