Features

• 🚀 Ready in minutes — install, sign in, and start chatting.
• 🔀 Native model picker — choose your models directly in Copilot Chat.
• 🧠 Reasoning built in — pick the thinking effort and follow reasoning live.
• 📊 Quota at a glance — see remaining usage and redeem available Codex resets with confirmation.
• ✨ More than chat — generate commit messages, titles, and summaries.
• 👥 Account load balancing — add multiple accounts for the same provider; CLIProxyAPI distributes requests round-robin across available accounts and skips those in cooldown.

[!WARNING] Use entirely at your own risk and discretion. This extension routes chat through your personal AI subscription accounts over OAuth. Accessing subscriptions outside their official apps may violate provider terms and could result in rate limiting or account suspension.

Demo

Quick start

Requires VS Code 1.131+ and the GitHub Copilot Chat extension.

1. Install — get Universal Chat Provider from the VS Code Marketplace. Prefer to build it yourself? See Development.
2. Add an account — accept the Add Account prompt (or run Universal Chat Provider: Add Account), pick a provider, and complete OAuth in your browser. Models refresh automatically.
3. Chat — open Copilot Chat and select a model under Universal Chat Provider.

Manage everything from the status bar item or the Universal Chat Provider: Manage Provider command — inspect quota, redeem OpenAI-provided Codex reset credits, list/remove accounts, restart, update, or reset the managed server.

Commit messages, titles, and summaries

Let your models handle Copilot's light background tasks and Explore searches. Run Universal Chat Provider: Set Utility Model (or use the status bar menu), pick a model (and thinking effort), done. A small, fast, inexpensive model is usually the best fit. Clear it to undo.

How it works

The command points Copilot's chat.utilityModel, chat.utilitySmallModel, and chat.exploreAgent.defaultModel settings at your selected model, so commit messages, chat titles, summaries, and Explore searches run through your accounts. It also sets chat.byokUtilityModelDefault to mainAgent, so remaining utility flows fall back to your selected chat model instead of Copilot's. When the model supports thinking levels, the command also asks for the utility Thinking Effort; commit messages use chat.utilitySmallModel plus that effort. No Copilot subscription required.

How it works

The extension runs CLIProxyAPI locally and registers its models with VS Code. Your subscriptions then appear directly in the Copilot Chat model picker.

Reference

Settings

Key	Description	Type	Default
▿ Connection
universalChatProvider.server.mode	How CLIProxyAPI is provided.	string	"managed"
universalChatProvider.baseUrl	CLIProxyAPI server URL. Used only in external mode.	string	"http://127.0.0.1:8317"
universalChatProvider.server.managementKey	External mode only. Enter your server's existing management key, not its bcrypt hash. Managed mode generates and stores its own key automatically and ignores this setting.	string	""
▿ Managed Server
universalChatProvider.server.updatePolicy	How managed CLIProxyAPI updates are handled.	string	"automatic"
universalChatProvider.server.version	CLIProxyAPI release used when update policy is Manual. Use latest or an exact version (8.0.3 or newer).	string	"latest"
universalChatProvider.server.extraConfig	Custom YAML overrides for the managed server, such as proxy, retries, and account selection. Overriding connection settings, credentials, or providers can break the extension.	string	"# requests:\n# proxy-url: http://127.0.0.1:7890\n# routing:\n# strategy: fill-first\n# retry:\n# request-retry: 3"
▿ Status Bar
universalChatProvider.showQuotaWarnings	Warn in the status bar when the model in use is low on quota.	boolean	true
universalChatProvider.quotaWarningThreshold	Remaining-quota percent below which the status bar warning appears.	number	10
▿ Provider Specific
universalChatProvider.codex.webSearch	Let Codex look things up online and open web pages while it answers.	boolean	false
universalChatProvider.claude.webSearch	Let Claude look things up online while it answers.	boolean	false
▿ Advanced
universalChatProvider.debugLevel	Diagnostic detail to collect. Request Logging prompts before restarting the managed server and writes sensitive payloads to disk.	string	"off"

Commands

Command	Title
universalChatProvider.manage	Universal Chat Provider: Manage Provider
universalChatProvider.login	Universal Chat Provider: Add Account (Login)
universalChatProvider.manageAccounts	Universal Chat Provider: Manage Accounts
universalChatProvider.showQuota	Universal Chat Provider: Show Quota
universalChatProvider.restartServer	Universal Chat Provider: Restart Managed Server
universalChatProvider.updateBinary	Universal Chat Provider: Update Proxy Binary
universalChatProvider.resetServer	Universal Chat Provider: Reset Managed Server
universalChatProvider.configure	Universal Chat Provider: Configure Connection
universalChatProvider.refresh	Universal Chat Provider: Refresh Models
universalChatProvider.setUtilityModel	Universal Chat Provider: Set Utility Model (commit messages, chat titles, summaries)
universalChatProvider.clearCredentials	Universal Chat Provider: Clear Stored API Key
universalChatProvider.showLogs	Universal Chat Provider: Show Logs
universalChatProvider.showServerLogs	Universal Chat Provider: Show Server Output
universalChatProvider.openSettings	Universal Chat Provider: Open Settings

Advanced

Session identity

The extension requires CLIProxyAPI 8.0.3 or newer. The extension sends a model-scoped prompt_cache_key derived from Copilot's conversation ID, with an opening-message fingerprint fallback. It does not also send X-Session-ID to the proxy.

Endpoints added through the extension receive these CLIProxyAPI header defaults, matched by exact hostname:

The proxy resolves $CPA-SESSION-ID from its session identity. Existing managed endpoints receive missing defaults when their config is regenerated. Existing headers take precedence regardless of casing, including empty values; server.extraConfig remains the final override. Other endpoints are unchanged.

Existing external-server endpoints are not rewritten. Add the relevant mapping under that provider's headers in your CLIProxyAPI configuration, or re-add the endpoint through the extension. Migrate custom mappings that copy $X-Session-ID to $CPA-SESSION-ID, since the extension no longer supplies the former header. Update older pinned or external proxies before relying on this behavior. A one-time session/cache reset may occur when changing identity sources.

Extra managed configuration

Set universalChatProvider.server.extraConfig to a CLIProxyAPI v8 YAML mapping in the multiline settings field. For example:

routing:
  strategy: fill-first
  retry:
    request-retry: 5

Mappings merge recursively; other values replace generated values, and invalid YAML is rejected before a requested restart stops the server.

Managed-server updates

Updates download automatically by default and restart the managed server as soon as no requests are active. Change universalChatProvider.server.updatePolicy to suggestUpdates or manual if you'd rather review or pin them yourself. Manual mode uses the configured version, which defaults to latest; enter an exact version (8.0.3 or newer) to pin it. An installed release older than 8.0.3 is replaced with a supported one regardless of the update policy.

Bring your own CLIProxyAPI server

Prefer to run CLIProxyAPI yourself, such as on a remote or shared machine?

1. Set universalChatProvider.server.mode to external.
2. Start CLIProxyAPI and complete the provider login there.
3. Use the Import API Key notification action when a local config is found, or run Configure Connection to enter the URL and key manually.

The server must run CLIProxyAPI 8.0.3 or newer. Adding or removing an OpenAI-compatible endpoint through the extension saves the server's config.yaml in the v8 layout; CLIProxyAPI keeps unrecognized sections as comments.

The API key is stored in VS Code SecretStorage. In external mode, the extension never starts or stops the server. To use Add Account and Manage Accounts, enter your server's existing management key, not its bcrypt hash, in universalChatProvider.server.managementKey. Managed mode generates and stores its own key automatically and ignores this setting.

Development

pnpm install
pnpm vscode:dts
pnpm check          # lint + typecheck + tests + build
pnpm ext:package    # produce an installable .vsix

Press F5 from VS Code Insiders to launch the Extension Development Host with the proposed APIs enabled.

License

MIT · Not affiliated with GitHub, OpenAI, Anthropic, or Google.