# Connect VS Code

> Add the Flowkiwi MCP server to VS Code and use it from GitHub Copilot chat.

This page connects VS Code (GitHub Copilot chat) to the MCP server of a Flowkiwi instance. Read [Connect an AI agent (MCP)](/mcp/) first for the sign-in, access levels and troubleshooting.

Sign in and pick your instance to fill your server URL into the snippets below (optional, the sign-in only serves the docs):

## Add the server

### In the workspace

Create or edit `.vscode/mcp.json` in your workspace. VS Code lists servers under `servers`:

```json title=".vscode/mcp.json"
{
  "servers": {
    "flowkiwi": {
      "type": "http",
      "url": "https://<tenant>.product-management.flowkiwi.net/api/mcp",
      "headers": {
        "Flowkiwi-Api-Version": "2026-07"
      }
    }
  }
}
```

The `headers` block is optional; it pins the [API version](/mcp/#choose-the-api-version). The file holds no secret, so you can commit it: each person signs in with their own account.

### For all your workspaces

Run **MCP: Open User Configuration** from the Command Palette and add the same `flowkiwi` entry under `servers`. You can also run **MCP: Add Server** for a guided setup: choose an HTTP server and paste the URL.

> **No OAuth settings to add**
>
> VS Code identifies itself to Flowkiwi on its own, so the `oauth` setting is not needed. Leave it out.

## Sign in

On the first connection, VS Code opens your browser on the Flowkiwi sign-in. Sign in, pick the organization and the access level on the consent screen, and go back to VS Code.

## Use it

Open Copilot chat in agent mode: the Flowkiwi tools are available, and the **Configure Tools** button lets you turn each of them on or off. Ask, for example: "List the draft products of the Kitchen category". Depending on your settings, VS Code asks you to confirm each tool call.

> **Several instances**
>
> Access is bound to one instance URL. Add one entry per instance, with a distinct name such as `flowkiwi-acme` and `flowkiwi-acme-staging`.