Let Claude Build Your Slides: Preso's MCP Server for AI Agents

You sit down to build a pitch deck and stare at a blank slide. The clock ticks while you align text boxes, hunt for the right chart, and wonder if the font even matches your brand. You already know the story. The deck should be the easy part.

Claude can write the story. Preso can design the deck. With Preso's MCP server, the two talk to each other. You describe the deck you need in a prompt, and Claude calls Preso to generate a finished, on-brand presentation. No more blank slides. No more pulling your hair out over alignment.

This guide walks you through connecting Claude and Preso, so an AI agent does the heavy lifting and you ship a beautiful deck in minutes.

Prerequisites

Before you start, gather these pieces. The setup is straightforward, but you need the right accounts and a few minutes of configuration.

• A Preso account. Sign up at Preso's pricing page. You will need at least the plan that supports API access, which is where the MCP server lives. If you are building for a team, the SaaS & Startups decks page outlines how startups use Preso for investor and board decks. Check it to make sure your plan aligns with your volume needs.
• A Claude Desktop or API key. You can use the Claude Desktop app (available for macOS/Windows) or access Claude via the Anthropic API. For this guide, we will use Claude Desktop because it connects to MCP servers directly. The official Tool Use with Claude - Anthropic Developer Documentation explains the underlying mechanism that makes this work.
• Your brand assets. Preso builds decks that actually look like your brand. Gather your brand colors, fonts, logo, and any specific imagery or template references. You will feed these into the prompt.
• Basic familiarity with JSON configuration. You will need to edit a configuration file for Claude Desktop. If you have ever tweaked a claude_desktop_config.json file, you are good. If not, do not worry. We will walk through it.

Step 1: Understand How MCP Turns Claude into a Deck Builder

The Model Context Protocol (MCP) is an open standard that lets AI models discover and use tools. Think of it as USB-C for AI agents. Instead of building a custom integration for every data source or action, you plug an AI model into an MCP-compatible server and the model learns what the server can do.

Anthropic introduced MCP and documented its design in the Model Context Protocol (MCP) - Official Introduction. That introduction is worth a read if you want to understand the protocol's architecture. For our purpose, the important part is this: Claude can call external tools via MCP. One such tool is Preso's presentation generation engine.

When you connect Preso's MCP server to Claude, the agent gains the ability to create slides. The flow works like this:

1. You prompt Claude with a deck request.
2. Claude interprets the story, structure, and brand details.
3. Claude calls the Preso tool, passing along the slide content, design preferences, and brand assets.
4. The Preso MCP server returns a fully designed, editable deck — usually in a shareable link or directly downloadable as a PowerPoint or Google Slides file.

This is not a simple template fill. Preso's headless presentation engine leverages the same AI design pipeline that powers its interactive builder. The result is a deck that respects visual hierarchy, data visualization, and narrative flow, not just a title-and-bullet placeholder.

Pro tip: Know the difference between a text summary and a structured deck. Claude can write a 10-slide outline as a markdown list. But without a tool, that outline is just text. With the Preso MCP server, that outline transforms into designed slides with branded layouts, charts, and imagery. The tool call turns intent into an actual presentation file.

Step 2: Set Up Preso's MCP Server

Preso's MCP server ships as a standard MCP-compatible server that you run alongside your AI agent. It is the same engine that powers Generate decks from your stack — so every slide it produces is consistent with the quality you get from the Preso app.

Here is how to get it running:

2.1. Get Your Preso API Key

Log in to your Preso account and navigate to the API page. If you do not yet have an API key, request one. The API key authenticates the MCP server so it can call Preso's generation endpoint on your behalf. Keep this key secure; it will live in your configuration file.

2.2. Clone or Download the MCP Server

Preso maintains an open-source MCP server repository. You can find it on GitHub. (If you are following along with a similar architecture, the anthropics/claude-mcp-slides - GitHub Repository provides a reference implementation that inspired Preso's approach.) Clone the Preso MCP server repository:

git clone https://github.com/trypreso/mcp-server.git
cd mcp-server

Warning: Always check the repository's README for the latest dependencies and environment requirements. The server requires Node.js 18+ and an active internet connection to reach Preso's API. If you are behind a corporate proxy, configure your environment accordingly.

2.3. Install Dependencies and Configure the Environment

Install the Node.js dependencies and set your Preso API key as an environment variable:

npm install

Create a .env file in the root of the repository:

PRESO_API_KEY=your-api-key-here

When the server starts, it reads this key and makes it available to the tool definitions.

2.4. Start the Server Locally

Run the server with:

npm start

The server listens on a local port (default: 3000). You will point Claude Desktop to this port.

If you prefer to run the server in a container, the repository includes a Dockerfile. That option is ideal for team or cloud deployments where you want the MCP server always available.

Step 3: Connect Claude Desktop to Preso's MCP Server

Claude Desktop connects to MCP servers through a configuration file. Open your Claude Desktop settings and locate the claude_desktop_config.json file. On macOS, it is usually at ~/Library/Application Support/Claude/claude_desktop_config.json.

Add an entry under mcpServers:

{
  "mcpServers": {
    "preso": {
      "command": "node",
      "args": ["/path/to/your/mcp-server/index.js"],
      "env": {
        "PRESO_API_KEY": "your-api-key-here"
      }
    }
  }
}

The preso key is an arbitrary label. It will appear in Claude's tool list. Restart Claude Desktop to load the new configuration.

After restart, open Claude and type "What tools do you have?" It should list the Preso tools, including generate_presentation, along with a description of the parameters it needs.

Pro tip: Keep your brand guidelines nearby. When you prompt Claude, include your brand details in the system message or at the top of the conversation. Because the tool call carries that context to Preso, the generated deck will respect your fonts, colors, and logo placement from the very first draft. Bookmark the Many designs for one deck page to see how Preso can iterate across different looks while staying on-brand.

The Integrating MCP with Claude: A New Era for AI Agents blog post on Anthropic's official site explains the broader vision. While that post focuses on data connectors, the same mechanism works for any tool — including a slide builder.

Step 4: Craft a Prompt That Produces a Polished Deck

The prompt is where your deck takes shape. A vague request like "make a pitch deck" will produce a vague deck. But a prompt that defines the audience, the structure, the brand, and the desired output format gives Claude everything it needs to call the tool intelligently.

Here is a template that works well:

You are my presentation design assistant. I need an investor pitch deck for my startup, Nymble, a B2B analytics platform for quick-service restaurants. The deck should have 10 slides: title, problem, solution, market size, product demo, traction, business model, competitive landscape, team, and ask. Use a clean, modern aesthetic with my brand colors (#0047AB and #FFFFFF), the font Inter, and my logo attached below. The deck must open with a hook that frames the restaurant data blind spot. Include a chart for market size and a comparison table for competitors. Generate the deck using the Preso tool and give me a shareable link.

When you send this, Claude calls the generate_presentation tool with a structured payload. The payload includes the slide content, design tokens, and any uploaded assets. Preso processes this in seconds and returns a link to a fully designed deck.

Notice what makes this prompt effective:

• Specific slide count and titles. Claude organizes the content accordingly.
• Visual instructions. Defining a color palette and font prevents a generic default.
• Data visualization requests. Instead of just "add a chart," you say what chart and for what data.
• Tone direction. "Hook that frames the restaurant data blind spot" guides the AI to write a compelling intro.

Warning: Do not confuse the prompt with the slide content itself. Your prompt is the instruction for the AI, not the finished copy. If you want precise wording on slides, include that in the prompt or attach a text file. The model will faithfully relay the content to Preso.

If you are building a sales deck, tailor the prompt like this: "Pull the account details, describe the angle, and design a personalized, on-brand pitch." The Sales & Revenue decks page outlines this use case in detail.

Step 5: Customize the Output with Brand Assets and Variations

One of the biggest complaints with AI-generated slides is that they look generic. Preso solves this by accepting brand assets directly through the API call. When Claude passes these to the MCP server, the engine applies them as design constraints.

Here is what you can control:

• Logo: Upload an SVG or PNG via the prompt (attach it in the conversation) or reference a URL.
• Color palette: Specify hex codes for primary, secondary, and accent colors.
• Font family: Preso supports Google Fonts and many system fonts. Name your font and it will be used across titles and body text.
• Image style: You can describe a visual direction ("dark gradient backgrounds with neon accents") and Preso's AI image generation will produce matching visuals.
• Layout preferences: Indicate if you prefer a text-heavy slide, a data-focused layout, or a bold visual statement.

The Plain English to a beautiful deck feature page shows how Preso turns a short description into a polished deck. The MCP server extends this so Claude does the describing on your behalf.

After the first deck is generated, you can ask Claude to iterate:

The deck looks great, but can you make the competitive landscape slide more visual and use a radar chart instead of a table?

Claude will call the tool again with updated parameters. You can also generate multiple design directions for the same content just by tweaking the style description in your prompt. That way, you walk into a presentation review with options.

Step 6: Iterate and Edit the Deck — You Are Still in Control

The deck Preso generates is not a static PDF. It is an editable presentation that you can open in the Preso editor, PowerPoint, or Google Slides. That matters because no AI gets every detail perfect on the first shot. You may want to tweak a headline, swap an image, or add a custom data visualization.

When you receive the shareable link from the MCP server, you can:

• Open the deck in Preso's editor and adjust slides with an AI assistant that speaks your deck's context. The A real story, in any language page highlights how Preso writes a narrative that you can then edit.
• Download a .pptx file and continue in PowerPoint. All elements remain native shapes and text boxes, so you are not stuck with a weird import.
• Export to Google Slides for team collaboration.
• Use Preso's Decks that present themselves feature to add an AI voice-over and turn the deck into a self-running video.

Pro tip: Treat the generated deck as a strong first draft. In traditional presentation building, the first draft often takes hours. With this flow, you get to the 90th percentile in minutes. Spend your remaining time on the details that require human judgment: adjusting the story arc for a specific investor, refining the language so it feels like you, and rehearsing the delivery.

The editing step also makes this workflow safe for enterprise teams. You can generate a deck programmatically, then have a human review every slide before it goes out. Preso's analytics (check Turn numbers into slides that land for how engagement tracking works) tell you which slides hold attention, so you know where to focus your edits.

Step 7: Share, Secure, and Export the Deck

Once you have a deck you are proud of, you need to get it into the right hands. Preso gives you multiple distribution options, all reachable after the MCP generation:

• Share a web link. The deck lives on Preso's platform with a unique URL. You can control access with password protection and view-only or comment permissions.
• Export to PowerPoint or Google Slides. If your workflow requires .pptx or Google Slides, a single click converts the deck with high fidelity.
• Embed the deck in a Notion page, website, or CRM. This is powerful for sales teams who want to track deck engagement within their existing tools.

If you are building a product that generates decks for users, the Headless presentations page shows how businesses integrate Preso's API to serve branded decks without ever touching the editor. The MCP server is the latest layer in that headless story, letting AI agents be the trigger.

For founders who use Claude extensively, you can save the generated deck to your Preso workspace and reuse it as a template for future fundraises. The SaaS & Startups decks page is full of examples where one good deck becomes a repeatable asset.

Why This Workflow Matters for AI Builders and Teams

The whole point of this setup is speed without sacrificing brand quality. Founders who used to spend a weekend building an investor deck can now have a conversation with Claude on Friday afternoon and walk into a Monday meeting with a polished presentation. Sales teams that personalize each pitch can feed account data into a prompt and get a tailored deck for every prospect.

The technical backbone is real, not hype. Research like the paper Autonomous AI Agents: The Role of Standardized Protocols underscores why protocols like MCP are essential: they make tool adoption repeatable across models. You are not locked into a single AI provider. Claude uses MCP today, and other models will follow the same standard. Your investment in a well-configured MCP server pays forward.

Coverage from Wired also highlights how MCP is reshaping enterprise automation. For presentations, that means moving from a human-does-everything model to a human-in-the-loop model where the AI assembles the deck and the human refines the strategy.

Step 8: Take It Further with Webhooks and Automation

If you are a developer, you can chain Preso's MCP server into larger automated workflows. The API | Preso page details how to set up webhooks so that when a deck is generated, your system gets notified. For example:

• A CRM update triggers a new sales deck via a webhook payload.
• A scheduled job runs every Monday morning to pull the latest KPIs and generate a weekly report deck through the MCP server.
• A data pipeline that monitors competitor activity triggers a competitive update deck for the strategy team.

Because the MCP server is just a thin wrapper around Preso's REST API, you can call it directly from your own service or script. The Generate decks from your stack feature page includes examples of how teams automate deck creation from databases, Notion, Airtable, and analytics tools.

Pro tip: Start simple, then automate. The most common mistake is trying to automate everything on day one. First, get comfortable generating decks through Claude conversations. Once you trust the output, hook the MCP server into a Slack command or a Zapier flow. That way, you learn what parameters consistently produce great results before scaling.

Summary and Key Takeaways

You came into this article staring at a blank slide. By now, you have a way to never stare at it again. Here is what you should take away:

• The blank slide problem is a tool gap, not a creativity gap. You have the story. MCP connects your story to a design engine that builds the deck.
• Preso's MCP server is a standard MCP tool that any compatible AI agent can use. Claude Desktop works out of the box, and the same server will work with other models that adopt MCP.
• Set up is simple: clone the server, configure your API key, add it to Claude's config file, and start prompting.
• The prompt is the brief. Be specific about slides, brand, visuals, and data. The more you give it, the more on-target the deck.
• You remain in control. The output is an editable file. You can iterate via the tool, edit in Preso, PowerPoint, or Google Slides, and add voice-over or analytics.
• This scales. From a single founder to a sales team to a product that generates decks for customers, the same MCP server and API handle it.

Build your next deck with Preso. Describe the story, and let Claude turn it into a designed presentation. Got a deck you need to ship next week? Start a conversation with Claude, connect Preso's MCP server, and watch the slides come together.