Connect VS Code

VS Code's built-in MCP support (GA in 1.99+) and Continue.dev both speak the MCP OAuth flow. Once connected, the AI side panel can read and modify your Moly workspace.

Before you start

You'll need:

• VS Code 1.99 or newer
• Either Copilot Chat (with MCP enabled) or the Continue.dev extension
• A Moly account

Copilot Chat (built-in MCP)

1. Open Command Palette → MCP: Add Server.
2. Choose HTTP as the transport.
3. Set:
   • Server name: moly
   • URL: https://api.moly.so/mcp
4. VS Code opens the Moly login in your browser. Sign in.
5. Back in VS Code, the server shows as connected in the MCP panel.

You can also add it via .vscode/mcp.json:

{
  "servers": {
    "moly": {
      "url": "https://api.moly.so/mcp",
      "type": "http"
    }
  }
}

VS Code stores the OAuth tokens in the system keychain. Don't put credentials in the JSON.

Continue.dev

1. Open Continue → Configure MCP servers.
2. Add an entry to ~/.continue/config.json:

{
  "experimental": {
    "modelContextProtocolServers": [
      {
        "transport": {
          "type": "http",
          "url": "https://api.moly.so/mcp"
        }
      }
    ]
  }
}

3. Reload Continue (Continue: Reload Configuration from the command palette). It opens the OAuth flow on first call.

Verify

In Copilot Chat or Continue.dev:

Show me tasks assigned to me this week.

The assistant should call search_tasks against Moly and stream the result.

Revoking access

Moly side: Settings → Integrations → Connected MCP clients → Disconnect the relevant row.

VS Code side: depends on the client.

• Copilot Chat: MCP: Remove Server from command palette.
• Continue.dev: remove the entry from config.json.

Workspace vs user config

A user-level mcp.json applies everywhere. A workspace-level .vscode/mcp.json only applies to that project — useful if different projects use different Moly workspaces.