# Continue in VS Code

Source: [https://docs.aventail.co.uk/docs/integration/vscode](https://docs.aventail.co.uk/docs/integration/vscode)

First [test the local API](https://docs.aventail.co.uk/docs/getting-started/run-first-interface) and leave your model serving.

### Connect Continue to Link

Screenshots from the original guide. Use your own model ID and port.

1.  **Install Continue**
    
    In VS Code, open Extensions, search for Continue and install the extension by Continue.
    
    [Screenshot: Continue search result and Install controls — close-up from Continue in VS Code.](https://docs.aventail.co.uk/img/guides/context/continue.png)
    
2.  **Open local configuration**
    
    Open Continue. If the setup screen appears, choose Skip and configure manually. Otherwise, open the local YAML configuration from settings.
    
    This setup screen also offers Ollama; use manual configuration for the Link endpoint.
    
    [Screenshot: Continue setup panel with Skip and configure manually highlighted.](https://docs.aventail.co.uk/img/vscodestep_4_.png)
    
3.  **Add your model**
    
    Use the complete YAML below. Set the exact model ID and an apiBase ending in /v1, then save.
    
    The screenshot uses Qwen3-LLM as its example model.
    
    [Screenshot: Continue YAML showing name, version, schema v1 and an OpenAI-compatible model with apiBase.](https://docs.aventail.co.uk/img/vscodestep_5_.png)
    
4.  **Send a test prompt**
    
    Select your configured model in Continue and send a short message. Keep Link serving while you use it.
    
    [Screenshot: Continue chat composer with a selected local model and a test prompt.](https://docs.aventail.co.uk/img/vscode4_.png)
    
5.  **Inspect the response**
    
    Wait for a complete answer. If the request fails, check the model ID, endpoint and where the extension is running.
    
    This original capture shows reasoning in progress; wait for the final answer.
    
    [Screenshot: Continue chat panel showing generated reasoning while a response is in progress.](https://docs.aventail.co.uk/img/vscodestep_7_.png)
    

**Read the Continue setup and copy YAML**

First [test your local model](https://docs.aventail.co.uk/docs/getting-started/run-first-interface). Keep it serving while you configure Continue.

## How the connection works

```mermaid
flowchart TB
  A["Continue in VS Code"] -->|"OpenAI-compatible request"| B["Link endpoint"]
  B --> C["Model on your device"]
  C -->|"Response"| A
```

## Configure Continue

Install the Continue extension in VS Code and open its local YAML configuration from Continue's settings. Add this model to your existing configuration; preserve any other providers you use.

A minimal complete configuration is:

```yaml
name: Local Aventail
version: 1.0.0
schema: v1
models:
  - name: Aventail chat
    provider: openai
    model: Gemma 4 (X-Small)
    apiBase: http://127.0.0.1:8100/v1
    apiKey: locai-local
    roles:
      - chat
```

Replace the port and model ID with the values from your [successful API test](https://docs.aventail.co.uk/docs/getting-started/run-first-interface). `locai-local` is a client placeholder for the tested unauthenticated local endpoint, not a Control account key. Use the configured credential if your endpoint requires authentication.

This uses Continue's [OpenAI-compatible provider](https://docs.continue.dev/customize/model-providers/top-level/openai) and [configuration schema](https://docs.continue.dev/reference).

## Test the connection

Save the configuration, select **Aventail chat** in Continue and send a short prompt. Confirm a complete answer appears.

If the model is missing, check YAML indentation and the required `name`, `version` and `schema` fields. For connection errors, confirm where the extension runs: in a remote VS Code session, `127.0.0.1` refers to that remote environment.

Start with chat. Autocomplete, editing and tools need separate configuration and a model that supports those capabilities.