# Automate with Claude & MCP

If you've used [Frase's built-in Agent](/get-started/meet-the-frase-agent.md), you already have a sense of what it looks like to describe a goal and have AI handle the execution. The MCP integration takes that further — it connects Frase directly to Claude, Anthropic's AI system, so you can operate the full Frase platform through a conversation in Claude Code.

This article is for people who want to understand what that means in practice and how to get started — no technical background required.

### What MCP Actually Is (In Plain Language)

MCP stands for Model Context Protocol. It gives Claude the ability to use Frase the same way you do — running research, creating briefs, auditing your site, finding content gaps, optimizing articles — but driven entirely by a conversation.

Instead of opening Frase, navigating to the right tool, and working through a workflow manually, you open Claude and describe what you want. Claude figures out which Frase tools to use, runs them in the right sequence, and brings the results back to you automatically.

The result is a seamless workflow that feels less like operating software and more like briefing a colleague who then goes and does the work.

### What You Can Ask Claude to Do

Here are some examples of what this looks like in practice:

* *"Find my top ten content gaps and create a brief for each one."*
* *"Run a site audit and give me a plain-language summary of the most critical issues."*
* *"Research this keyword, build a brief, and write a draft optimized for AI visibility."*
* *"Which of my published articles are showing signs of content decay?"*
* *"Create a content cluster around email marketing — pillar page plus five supporting article briefs."*
* *"What should I publish next based on my current opportunities?"*

Each of these would normally involve opening multiple tabs, running several tools in sequence, and synthesizing the results yourself. With the MCP connection, it's one request.

### Getting Started With Claude Code

Claude Code is Anthropic's command-line tool that lets Claude work directly with tools and services on your behalf. It's the environment where the Frase MCP connection runs.

Here's what getting set up looks like, step by step.

{% stepper %}
{% step %}

### **Install Claude Code**

Claude Code runs in your computer's terminal. If you've never used a terminal before, don't let that put you off — the setup is straightforward and you only do it once.

Go to [claude.ai/code](https://claude.ai/code) and follow the installation instructions for your operating system. You'll need Node.js installed on your computer first. If you don't have it, the Claude Code setup page will point you to where to get it — it's a free download and takes about two minutes to install.
{% endstep %}

{% step %}

### **Connect the Frase MCP Server**

With Claude Code installed, run the following command in your terminal — replacing the placeholder at the end with your actual Frase API key:

```
claude mcp add frase -- npx -y @frase/mcp-server --api-key sk_live_your_api_key
```

To find your API key, go to **Settings → API Keys** inside Frase and copy it from there. Once you run this command, Claude Code will automatically start the Frase server whenever it's needed. You only need to do this once.
{% endstep %}

{% step %}

### **Start a Session**

Once the setup is complete, open Claude and start a conversation like you normally would. The Frase MCP server runs in the background automatically — there's nothing extra to launch or configure.

To confirm everything is connected, try this as your first message:

*"List my connected Frase sites."*

If Claude responds with your site names, you're ready to go.
{% endstep %}
{% endstepper %}

{% hint style="success" %}
**Need setup support?** We'd love to walk through this setup with you. Reach out to us at <support@frase.io> and we'll find a time to connect.
{% endhint %}

### What to Expect the First Time

The first session can feel a little different from using a regular chat interface. A few things worth knowing:

* **Claude will ask your permission before using each tool for the first time.** The first time Claude calls a Frase tool in a session, it will pause and ask whether you want to **Allow Once** or **Always Allow**. This is a safety feature built into Claude Code, not a sign that something is wrong.
  * Choosing **Allow Once** means Claude will ask again the next time it needs that same tool. Choosing **Always Allow** gives Claude permission to use that tool freely going forward without interrupting to ask.
* **Claude will tell you what it's doing.** As it works through a request, Claude describes the steps it's taking — which Frase tools it's calling, what it found, what it's doing next. This transparency is intentional. You're always in the loop.
* **Some requests take a moment.** When Claude is running a site audit or pulling research data, it's actually executing those workflows in Frase in real time. A request that involves multiple steps may take thirty seconds to a minute. That's normal.
* **You can ask follow-up questions.** The session is a conversation. If Claude surfaces ten content gaps and you want to go deeper on one of them, just ask. You can chain requests together naturally without starting over.
* **Claude won't make changes without telling you.** When Claude is about to do something consequential — creating a document, publishing content, running an audit — it will describe what it's about to do before it does it. You stay in control.

### A Few Things Worth Trying First

Once you're set up, here are some good starting requests to get a feel for what's possible:

1. Start with something low-stakes and informational — *"What are my top content opportunities right now?"* — to see how Claude interprets your data and presents results.
2. Then try a creation request — *"Research \[your keyword] and create a brief"* — to see the full research-to-brief workflow run end to end.
3. Then try something analytical — *"Which of my pages are ranking between position four and ten and are worth optimizing?"* — to see how Claude uses your Search Console data.

Each of these gives you a different window into what the integration can do. From there, the workflows you'll find most valuable will depend on your own content operation — what's most repetitive, what takes the most time, and where having a tireless collaborator would make the biggest difference.

### For Developers and Technical Teams

If you're a developer or want to explore the full technical capabilities of the Frase MCP server and API, the complete documentation is available here:

* [MCP Server](https://npmjs.com/package/@frase/mcp-server)
* [API Documentation](https://next.frase.io/api/docs)


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.frase.io/core-workflows/automate-with-claude-and-mcp.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.