# GitHub Copilot Chat with Claude through a gateway: Custom Endpoint setup, agent mode and fixes

VS Code's chat can run on models you bring yourself. The built-in Anthropic provider only accepts an API key, with no field for a different address, so a gateway is added through the Custom Endpoint provider. Choose the Messages API type and list the models in `chatLanguageModels.json`. It has been part of stable VS Code since version 1.122.

## What bring-your-own-key covers

| Feature | With your own model |
| --- | --- |
| Chat, ask and plan | Yes |
| Agent mode, built-in tools and MCP servers | Yes, if the model is marked `toolCalling: true` |
| Inline suggestions (code completions) and next edit suggestions | No: these still come from GitHub Copilot |
| Semantic search and other embedding-based features | No: these still need a GitHub account |

Since VS Code 1.122 these models work without signing in to GitHub and without a Copilot plan. On Copilot Business and Enterprise, BYOK is on by default, and administrators can turn it off with the **Bring Your Own Language Model Key** policy.

## Add the gateway as a Custom Endpoint

1. Open the chat view, open the model picker and choose **Manage Language Models**, or run **Chat: Manage Language Models** from the Command Palette.
2. Choose **Add Models**, then the **Custom Endpoint** provider.
3. Enter a display name and your gateway API key.
4. Select the API type **Messages**. VS Code then opens `chatLanguageModels.json` for the model list.

## chatLanguageModels.json

```json
[
  {
    "name": "AI Prime Tech",
    "vendor": "customendpoint",
    "apiKey": "${input:aiprimeApiKey}",
    "apiType": "messages",
    "models": [
      {
        "id": "claude-sonnet-5",
        "name": "Claude Sonnet 5",
        "url": "https://aiprimetech.io/v1/messages",
        "toolCalling": true,
        "vision": true,
        "maxInputTokens": 200000,
        "maxOutputTokens": 32000
      },
      {
        "id": "claude-opus-5-5",
        "name": "Claude Opus 5.5",
        "url": "https://aiprimetech.io/v1/messages",
        "toolCalling": true,
        "vision": true,
        "maxInputTokens": 200000,
        "maxOutputTokens": 32000
      },
      {
        "id": "claude-haiku-4-5",
        "name": "Claude Haiku 4.5",
        "url": "https://aiprimetech.io/v1/messages",
        "toolCalling": true,
        "vision": true,
        "maxInputTokens": 168000,
        "maxOutputTokens": 32000
      }
    ]
  }
]
```

| Field | Why |
| --- | --- |
| `id` | The model ID sent to the gateway |
| `url` | The full endpoint. VS Code uses a URL ending in `/messages` as-is; give the full path to avoid guessing |
| `toolCalling: true` | Required for agent mode. A model without it doesn't appear for agents |
| `maxInputTokens` + `maxOutputTokens` | Their sum must not exceed the model's context window |
| `apiKey: ${input:...}` | VS Code asks for the key once and stores it, so it isn't written into the file |

With the Messages type, VS Code sends the key in the `x-api-key` header, which is what the gateway expects. Model IDs must match the gateway's exactly.

## GPT models on the same key

The Custom Endpoint provider also speaks the OpenAI formats. To add GPT models from a key in the gateway's Codex group, add a second provider with `"apiType": "chat-completions"` and model URLs of `https://aiprimetech.io/v1/chat/completions`. A key reaches only the models of its own group, so Claude and GPT usually need separate entries with separate keys.

The older `github.copilot.chat.customOAIModels` setting and the **OpenAI Compatible** provider are deprecated in favour of Custom Endpoint. Move old entries into `chatLanguageModels.json`.

## Troubleshooting

| What you see | Cause | Fix |
| --- | --- | --- |
| The model doesn't appear in the picker | Needs a reload, or `toolCalling` is false, or the workspace is in Restricted Mode | Restart VS Code, set `toolCalling: true`, trust the workspace |
| The provider list has no Custom Endpoint | VS Code older than 1.122, or BYOK disabled by your organisation | Update VS Code; ask the admin about the BYOK policy |
| `401` / invalid API key | The stored key is wrong or belongs to another endpoint | Gear icon next to the provider, then **Update API Key** |
| Model not supported | The `id` doesn't match a gateway model ID | Use an ID from the table below |
| Context or token errors | `maxInputTokens` plus `maxOutputTokens` exceeds the window | Lower `maxInputTokens` |
| `Concurrency limit exceeded` | Several agents or windows running at once on one key | Run fewer in parallel |

For details, run **Developer: Set Log Level**, choose **Trace**, then open **Output** and pick **GitHub Copilot Chat**. The Chat Debug view shows the exact requests. A log line saying BYOK providers were unregistered because the policy disallows them points at an organisation policy.

| Model ID | Use it for |
| --- | --- |
| `claude-sonnet-5` | Default for coding and agent work: fast, strong, mid-priced. |
| `claude-opus-5-5` | Harder reasoning, large refactors, long agent runs. |
| `claude-fable-5-1` | Most capable and most expensive; save it for the hardest tasks. |
| `claude-haiku-4-5` | Cheap and quick: titles, summaries, classification, small steps. |

Prefer the terminal? The same key runs [Claude Code in VS Code](/claude-code/vscode/) through its own extension.

## FAQ

### Can I use my own Claude API key in GitHub Copilot?

Yes. In VS Code's chat choose Manage Language Models, Add Models, then the Anthropic provider for Anthropic's own API, or Custom Endpoint with API type Messages for a gateway.

### Can I change the Anthropic base URL in Copilot?

Not in the built-in Anthropic provider, which only takes a key. Use the Custom Endpoint provider and set each model's url to the gateway's /v1/messages endpoint.

### Do BYOK models work in Copilot agent mode?

Yes, when the model is marked toolCalling: true. Built-in tools and MCP servers keep working; inline code completions still come from Copilot.

### Do I need a Copilot subscription to use my own model?

Not since VS Code 1.122: BYOK models work without a GitHub sign-in or Copilot plan. Business and Enterprise organisations can disable it by policy.

---

Source: https://aiprimetech.io/learn/use-claude-api-with-github-copilot/
Provider: AI Prime Tech — independent Anthropic-compatible API gateway (not affiliated with Anthropic).