> ## Documentation Index
> Fetch the complete documentation index at: https://www.voiceflow.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Set up your AI agent

> Connect your AI assistant to Voiceflow, give it access to the docs and skills, and verify your first connection.

Use Claude Code, Codex, Cursor, or another supported assistant to work with Voiceflow in plain language. The setup prompt connects your account, makes documentation available, and installs Voiceflow skills where your client supports them.

## Before you begin

- Have a [Voiceflow account](https://creator.voiceflow.com/signup) and access to the workspace you want to use.
- Open your AI assistant. For local installation, run it on the computer where you want to use Voiceflow.

## Connect your assistant

Choose **Setup prompt** for guided setup of your Voiceflow connection, documentation access, and skills. Use **CLI** or your coding agent’s tab for manual connection instructions.

<Tabs>
  <Tab title="Setup prompt" icon="/docs/images/icons/message.svg">
    Paste this prompt into your AI assistant:

    ```text Setup prompt wrap
    Set up Voiceflow by following https://www.voiceflow.com/agents/setup
    ```

    Your assistant will reuse existing connections or set up CLI or MCP access, connect Voiceflow's documentation, and install the Voiceflow skills where your client supports them. Complete browser sign-in when prompted; your assistant will verify the connection before finishing.
  </Tab>
  <Tab title="Claude Code" icon="/docs/images/icons/claude.svg">
    <Steps>
      <Step title="Add Voiceflow">
        Run this command in your terminal:

        ```bash
        claude mcp add --transport http voiceflow https://mcp.voiceflow.com/mcp
        ```
      </Step>
      <Step title="Sign in to Voiceflow">
        Open Claude Code, run `/mcp`, select **voiceflow**, and choose **Authenticate**. Sign in to Voiceflow in your browser and approve the connection.

        Connecting grants your client access to read and change the projects your account can access. [MCP authentication](/docs/mcp/authentication) explains the permissions.
      </Step>
      <Step title="Confirm the connection">
        Ask Claude Code:

        ```text
        List my Voiceflow projects.
        ```

        A project list confirms the connection is working.
      </Step>
    </Steps>
  </Tab>
  <Tab title="Codex" icon="/docs/images/icons/open-ai.svg">
    <Steps>
      <Step title="Add Voiceflow">
        Run this command in your terminal:

        ```bash
        codex mcp add voiceflow --url https://mcp.voiceflow.com/mcp
        ```
      </Step>
      <Step title="Sign in to Voiceflow">
        If adding the server already completed sign-in, continue to the next step. Otherwise, run:

        ```bash
        codex mcp login voiceflow
        ```

        Sign in to Voiceflow in your browser and approve the connection.

        Connecting grants your client access to read and change the projects your account can access. [MCP authentication](/docs/mcp/authentication) explains the permissions.
      </Step>
      <Step title="Confirm the connection">
        Ask Codex:

        ```text
        List my Voiceflow projects.
        ```

        A project list confirms the connection is working.
      </Step>
    </Steps>
  </Tab>
  <Tab title="Cursor" icon="/docs/images/icons/cursor.svg">
    <Steps>
      <Step title="Add Voiceflow">
        Add this configuration to `.cursor/mcp.json` in your project, or merge the `voiceflow` entry into your existing `mcpServers`:

        ```json
        {
          "mcpServers": {
            "voiceflow": {
              "url": "https://mcp.voiceflow.com/mcp"
            }
          }
        }
        ```
      </Step>
      <Step title="Sign in to Voiceflow">
        When Cursor prompts you to authenticate, sign in to Voiceflow in your browser and approve the connection.

        Connecting grants your client access to read and change the projects your account can access. [MCP authentication](/docs/mcp/authentication) explains the permissions.
      </Step>
      <Step title="Confirm the connection">
        Ask Cursor:

        ```text
        List my Voiceflow projects.
        ```

        A project list confirms the connection is working.
      </Step>
    </Steps>
  </Tab>
  <Tab title="CLI" icon="/docs/images/icons/console.svg">
    Run these commands yourself in your local terminal.

    <Steps>
      <Step title="Install the Voiceflow CLI">
        On macOS or Linux:

        ```bash
        curl -fsSL https://raw.githubusercontent.com/voiceflow/cli/master/scripts/install.sh | bash
        ```

        On Windows, use PowerShell:

        ```powershell
        iwr -useb https://raw.githubusercontent.com/voiceflow/cli/master/scripts/install.ps1 | iex
        ```
      </Step>
      <Step title="Sign in to Voiceflow">
        ```bash
        vf auth login
        ```

        Follow the sign-in instructions in your terminal.
      </Step>
      <Step title="Confirm the connection">
        ```bash
        vf workspace list --output-format json
        ```

        A workspace list confirms the connection is working. If you use a personal or service key, access depends on that key's permissions; see [CLI authentication](/docs/cli/authentication).
      </Step>
    </Steps>
  </Tab>
</Tabs>

## What gets connected

| Connection | What it enables |
| --- | --- |
| Platform access | Read and manage Voiceflow resources through an existing connection, the CLI, or the platform MCP server. |
| Documentation | Find Voiceflow guidance through the docs MCP server, with a supported fallback when the host cannot connect it. |
| Voiceflow skills | Give supported assistants reusable guidance for setup and working with Voiceflow. |

Account access follows the permissions of your connected account or credential. Review [MCP authentication](/docs/mcp/authentication) or [CLI authentication](/docs/cli/authentication) for details.

## Try your first task

Start with a request that reads your project before making changes:

```text First task wrap
List my Voiceflow projects and ask me which one to use. Then summarize that agent’s instructions and knowledge sources without making changes.
```

Check that your assistant is using the project you selected. Once it can inspect the agent, ask it to help with a specific prompt, knowledge source, or workflow.

## If you get stuck

- **Sign-in did not finish:** retry the browser flow in your assistant. See [CLI authentication](/docs/cli/authentication) or [MCP authentication](/docs/mcp/authentication) for the connection you are using.
- **Tools do not appear after connecting:** follow your client’s setup guide: [Claude Code](/docs/mcp/claude-code), [Codex](/docs/mcp/codex), or [Cursor](/docs/mcp/cursor).
- **Skills are unavailable:** support depends on the app. Use the setup prompt so your assistant can check its actual capabilities and explain what it configured.

## Keep building

<CardGroup cols={2}>
  <Card title="Understand how agents work" icon="/docs/images/icons/agent.svg" href="/docs/documentation/build/how-agents-work" cta="Read the guide">
    Learn how instructions, playbooks, workflows, knowledge, and tools fit together.
  </Card>
  <Card title="Explore the platform MCP" icon="/docs/images/icons/mcp.svg" href="/docs/mcp/overview" cta="Explore MCP">
    See how your assistant can work with your Voiceflow account.
  </Card>
</CardGroup>