Unify Provider
Connect your ChatGPT subscription (Codex allowance) to VS Code's native Chat and Agent (Copilot Chat) through OpenAI's official Sign in with ChatGPT.
No API key, URL, port, separate backend, Node.js, or Codex CLI is needed.
Requirements
- VS Code 1.104 or later, running locally.
- GitHub Copilot Chat for the native Chat/Agent interface. Availability of bring-your-own-model (BYOK) features depends on your VS Code version and organization policies; current VS Code documentation does not require a Copilot subscription for BYOK models.
- A ChatGPT account eligible for Sign in with ChatGPT plan usage.
- Network access to auth.openai.com and api.openai.com, plus a browser for sign-in.
Getting started
- Install Unify Provider in VS Code.
- Open the Command Palette and run Unify Provider: Sign in with ChatGPT.
- Read the notice, choose Continue with ChatGPT, and complete sign-in and authorization in your browser. Agree to using your ChatGPT plan allowance; on first successful sign-in, acknowledge the You're using your ChatGPT plan confirmation with Got it.
- In the Chat model picker, select a model under Unify Provider (ChatGPT plan).
You must select a model once yourself; the extension cannot select it automatically. Models are discovered from your connected account, not a fixed list.
For the Copilot harness / Agents window (Agent Host), enable the experimental VS Code setting chat.agentHost.byokModels.enabled. This setting may change or be removed. The Local harness is a separate path. Unify Provider does not change your settings automatically.
Commands
Open the Command Palette to use these commands:
| Command |
Purpose |
| Unify Provider: Sign in with ChatGPT |
Connect your account, or sign in again after authorization expires or is revoked. |
| Unify Provider: Disconnect |
Disconnect and delete the saved session from VS Code Secret Storage. |
| Unify Provider: Manage Unify Provider |
Open account actions: sign in, disconnect, show status, show logs, or Manage usage. |
| Unify Provider: Show Status |
Show the current connection/account status and a Manage usage action. |
| Unify Provider: Show Logs |
Open the Unify Provider Output channel for safe diagnostics. |
Status bar
Click the ChatGPT status bar item to open account actions. Its text indicates:
- Using ChatGPT plan · account: connected to the displayed account. Remaining/total quota is shown only when provided by the service;
? means a value is unknown.
- ChatGPT: Connecting: sign-in is in progress.
- ChatGPT: Disconnected: no active connection; sign in to discover models.
- ChatGPT: Sign in again: authorization needs renewal.
- Usage limited: a plan or app usage limit was reached.
- Ineligible/access denied: account eligibility or authorization prevents access.
- Request failed or Status unavailable: check the logs and troubleshooting steps below.
The tooltip and account actions include Manage usage.
Troubleshooting
Models do not appear
Sign in first and check the status bar. Copilot Business/Enterprise administrators can disable BYOK through the Bring Your Own Language Model Key in VS Code policy. For the Copilot harness / Agents window, also check chat.agentHost.byokModels.enabled.
If models still do not appear, run Developer: Reload Window, then reopen the model picker. Older VS Code releases may have different BYOK account restrictions. When not signed in to GitHub, utility models may require separate configuration in VS Code (chat.utilityModel, chat.utilitySmallModel, or chat.byokUtilityModelDefault); selecting a main Chat model does not configure every auxiliary feature.
Requests fail or the network is unreliable
Run Unify Provider: Show Logs, or open View → Output and select Unify Provider. Check the diagnostic error and connection status before trying again.
Intermittent network_error errors can indicate a network, proxy, certificate, or gateway problem. When a response has no content-type, the extension automatically identifies its format from the body. If it still fails, check your system proxy and VS Code's http.proxy, http.proxySupport, and http.fetchAdditionalSupport settings. An HTTP 200 response alone does not guarantee a successful model response.
When reporting a problem, share only safe diagnostic metadata, not credentials, conversation text, or response bodies.
Usage limit or access denied
Open Manage usage to review your ChatGPT plan and app limits. Usage is shared with other apps using your plan. Limit and eligibility errors are not retried automatically, and reaching a limit does not switch to a paid API method. For access-denied errors, check account eligibility and the authorization you granted.
Sign in again
Run Unify Provider: Sign in with ChatGPT if authorization expired or was revoked. To replace a connected account, Disconnect first, then sign in again. You can also revoke this app's authorization in ChatGPT settings.
Privacy & data
- Credentials are stored in VS Code Secret Storage. Disconnect on shared computers to remove the saved session.
- Logs do not include tokens or conversation content; diagnostics use safe metadata and error identifiers.
- Model discovery and inference requests go directly to api.openai.com from the extension, with no third-party server operated by Unify Provider. Your selected chat context is sent to OpenAI to answer requests.
Limitations
- The extension cannot automatically select a model in the Chat picker.
- Inline completions, semantic search, and embedding-dependent features do not use this model connection.
- Image input is available only when the selected model supports it. Audio and non-text tool results are not supported.
- Required tool mode is not supported; requests requiring it are rejected rather than treated as automatic tool selection.
- Remote SSH OAuth callback forwarding is not supported. Use a local VS Code window; remote and Web environments are outside this extension's scope.
Disclaimer
Unify Provider is a free, locally running, MIT-licensed open-source third-party tool, not an OpenAI product and not endorsed by OpenAI. Sign in with ChatGPT (SIWC) is in preview: APIs, available models, eligibility, and policies may change. Account access cannot be guaranteed.
Requests count toward your ChatGPT plan's Codex allowance, shared with other apps. There is no extra charge from Unify Provider and no automatic paid fallback.
License
MIT.
