AI Manager

English | 简体中文 | 日本語

AI Manager is a VS Code desktop extension for managing OpenAI-compatible channels, dynamic model catalogs, and native VS Code Chat model bindings from one place.

Why AI Manager

In VS Code, AI features such as code completions and Git commit-message generation typically default to GitHub Copilot. Using another AI model often means manually changing settings in several places. AI Manager was built to bring model connections, management, and switching into one place, so you can choose the right model more freely for each workflow.

Requires VS Code 1.121.0 or later.

Highlights

• Manage OpenCode Go, OpenCode Console, and custom OpenAI-compatible channels.
• Refresh model catalogs while keeping the last successful cache available.
• Rename, filter, configure, and selectively enable discovered models.
• Register enabled models in the native VS Code Chat model picker.
• Bind separate models to Chat, Inline Chat, Plan, Plan implementation, Utility, and Utility Small settings.
• Share all channels, model catalogs, refresh state, and Chat bindings across local VS Code Profiles, with encrypted cross-device synchronization through Settings Sync.

Screenshots

Channel and model management

Chat model bindings

Quick Start

1. Open AI Manager from the Activity Bar.
2. Add a channel and enter its Base URL, endpoint paths, authentication mode, and optional API key.
3. Refresh the channel to load its model catalog.
4. Review the discovered models, adjust aliases or metadata when needed, and enable the models you want to use.
5. Open the Chat settings page in AI Manager and assign a channel and model to each setting you want to manage.

Newly discovered models are disabled by default.

Supported Protocols

Custom channels can configure independent endpoints for all four protocols and use Bearer, x-api-key, or x-goog-api-key authentication. A Gemini endpoint must include the {model} placeholder.

The protocol of each model is resolved in this order: explicit fields in the channel catalog, protocol metadata from models.dev, local rules for known model families, and finally the channel default. If a request fails because the endpoint for the selected protocol does not exist, AI Manager retries once with another configured protocol and remembers the result.

You can override the protocol and token limits from the model editor. Tool-calling capability comes from the model catalog: it is reported as declared when the catalog states it, and assumed available otherwise.

Cross-device Sync

All Profiles of the same VS Code edition share channels, the complete model catalog, refresh results, Chat bindings, and the encrypted vault through one local state directory. A Chat setting changed manually in one Profile is propagated to the others. Stable, Insiders, and compatible VS Code editions use separate directories.

When cross-device sync is enabled, the complete shared state is compressed, integrity-checked, and sent through VS Code Settings Sync. API keys are encrypted with PBKDF2-SHA256 and AES-256-GCM first. The master password is never saved; each Profile stores only its derived key locally in SecretStorage.

On another computer, sign in to the same VS Code account, enable Settings Sync, and unlock AI Manager once in every Profile that uses the extension. If the password is lost, the vault must be reset, which removes the synchronized ciphertext and API keys while retaining the local non-secret configuration.

Notes and Limitations

• The main Chat default model applies to new conversations and does not replace a model manually selected in an existing conversation.
• The Plan implementation model uses an experimental VS Code setting and may be unavailable because of the VS Code version or organization policy.
• Disabled, unavailable, or unconfigured models are not exposed to the native model picker.
• The extension reads and writes a local shared state directory, so it is declared as a UI extension. In Remote-SSH, WSL, and Codespaces windows it runs on the local machine rather than on the remote host.

Privacy

Request logs contain only the channel, alias, model ID, duration, HTTP status, and error category. AI Manager does not log prompts, responses, credentials, the sync master password, or derived keys, and it does not read local OpenCode authentication files.

When a model catalog is refreshed, AI Manager also requests https://models.dev/api.json to determine which wire protocol each model uses. That request carries no API key and no user data, and its result is cached for six hours.

Development

npm install
npm run check
npm run package:vsix

Use npm test for unit tests and npm run test:integration for the Extension Host integration test. Press F5 in VS Code to launch an Extension Development Host.