🖼️ Image Attachments

Attach files or inline base64 images to a query for vision-capable models.

The query tool accepts an optional attachments array for vision-capable models (Claude, GPT-5+).

📎 File attachment

Reads from disk:

{
  "type": "file",
  "path": "/absolute/path/to/screenshot.png"
}

Inline base64 data:

{
  "type": "blob",
  "data": "<base64-encoded-content>",
  "mimeType": "image/png"
}

Both attachment types support an optional displayName field.

🧰 Tools

The two tools exposed by the server — query and list_models — with example conversations.

⚙️ How It Works

How the server bridges your assistant and GitHub Copilot via the official SDK.