Maximizing Claude AI: Desktop App, MCP, Agents & Presentation Generation Guide

I don’t consider myself a developer! Although I’ve long understood the value of programming because it allows you to get to grips with reality on some digital issues, I do not consider myself an AI specialist either, but rather an AI user.

As a user attentive to technological trends, my curiosity is sometimes piqued by emerging innovations. Some disappear as quickly as they appeared, while others become unavoidable phenomena.

Faced with these novelties, I always ask myself the same question: “Can this be useful to me in any way?”

Why MCPs Deserve Your Attention in 2025

Model Control Protocols (MCPs) caught my attention through repetitive alerts from sources like Medium. I gradually understood that this topic could be important.

From my understanding, MCPs present an obvious benefit with the help of Claude Desktop for example. They offer the possibility of running language models like Claude Sonnet directly on your computer and designing functions that intervene on your machine like a super assistant.

There are numerous articles on the concept of MCPs, and I’ve tried to understand how they could impact my daily life.

Three Practical Applications for Non-Developers

Claude Desktop: AI Directly on Your Machine

Claude Desktop represents a great advancement for non-technical users, even though it still required some programming skills. This application allows you to access Claude Sonnet’s capabilities directly from your computer.

More on https://claude.ai/download

AI Agents: Automate Without Coding

AI agents constitute the second important application. These intelligent assistants can be configured to perform specific tasks on your computer, such as organizing files or extracting information, without you having to write a single line of code.

The Claude Extension for Google Sheets

The Claude extension for Google Sheets, accessible via an API key, transforms data analysis. It allows you to query your data in natural language and obtain relevant visualizations without in-depth knowledge of data analysis.

FAQ: MCPs for Non-Technicians

What exactly is an MCP?
A Model Control Protocol (MCP) is a set of rules that allow an AI model to interact with other systems, such as your computer or applications.

Do you need to be a developer to use these technologies?
No, that’s precisely the advantage of these advancements: they’re designed to be accessible to users without particular technical skills.

What is the simplest application to start with?
Claude Desktop is probably the most accessible entry point for non-developers wishing to explore these technologies.

Bonus: Generating Presentations with Claude

As a bonus, I’m also sharing a method for generating PowerPoint or Reveal.js presentations using Claude. I’ve had to do it and I’ll probably have to do it again. It’s simply a matter of generating a narrative path from a set of scattered notes and then, once this path has been structured by Claude, producing an output in PowerPoint or Reveal.js format!

This approach perfectly illustrates how non-developers can leverage advanced AI capabilities for everyday tasks, thus transforming their productivity without requiring in-depth technical skills.

1. Claude Desktop, MCP & Python

Nothing else to say other than what I said above, it is mainly for me the possibility of integrating functionalities which make functions interpreted by Claude Desktop interact, functions which can also have direct actions on my laptop.

MCP is an open protocol that standardizes how applications provide context to LLMs. Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect your devices to various peripherals and accessories, MCP provides a standardized way to connect AI models to different data sources and tools.

# the path to the config file for claude desktop
/Users/[username]/Library/Application Support/Claude/
# the config file for claude desktop
/Users/[username]/Library/Application Support/Claude/claude_desktop_config.json

The idea is simple and can be summed up by this dilemma:

  • Whitout MCP: Fragmented AI development
  • Whith MCP: Standardized AI development

The Model Context Protocol (MCP) lets you build servers that expose data and functionality to LLM applications in a secure, standardized way. Think of it like a web API, but specifically designed for LLM interactions. MCP servers can:

Expose data through Resources (think of these sort of like GET endpoints; they are used to load information into the LLM’s context)
Provide functionality through Tools (sort of like POST endpoints; they are used to execute code or otherwise produce a side effect)
Define interaction patterns through Prompts (reusable templates for LLM interactions)
And more!

Here is an extremely fast Python package and project manager, written in Rust and required to make the MCP server work. Read more on https://docs.astral.sh/uv/

See files in https://github.com/bflaven/ia_usages/tree/main/ia_using_claude/ia_claude_mcp_python

2. Create Agents in Claude

When you subscribe, you can make Claude work just for you by creating agents. This is bigger than just typing prompts – whether you’re looking for information, creative content, analysis, or technical help. It’s about turning Claude into your personal AI that follows your specific instructions and works with your content.

Defining an agent in Claude is done using three elements to complete: Name, Description, Set project instructions, and it’s possible to add content elements in “Project knowledge”.

Name:

SEO_AGENT_TEST_1

Description:

Test #1 SEO agent Claude

Set project instructions:

You will be my personal AI agent to help me complete various tasks related to marketing and content creation. This includes creating social media content, blog posts, ad copy, emails, and generating content that sounds like me.

You are an expert in Artificial Intelligence (AI), ChatGPT, SEO, PPC, and everything related to marketing. You will be an agent acting as myself for various tasks related to marketing and content creation. In all of your outputs, please follow the provided instructions:

Use the uploaded knowledge to mimic my writing style for all your responses
Use a writing tone that shows expertise, seriousness, and empathy; this should sound like it was written by a human
Write at an eighth grade reading level that's understandable to the average person
Do not use the following words or phrases in any output: "delve, tapestry, vibrant, landscape, realm, embark, excels, vital, comprehensive, intricate, pivotal, moreover, arguably, notably, dive into, intriguing, holistic, sail into the future, ethical considerations, area, realm, in the field of, make the world better, in the future, in essence.

1. voice matching
- Always mirror my conversational, straightforward writing style
- Use contractions naturally (I'm, You'll, we're etc.)
- Keep sentences punchy and direct
- Avoid formal and Academic language unless specifically requested
- Write in first person when creating content from my perspective 


2. readability requirements
- Target an eighth grade reading level consistently
- Use short paragraphs (2 or 4 sentences maximum)
- Break up text with subheadings every 2-3 paragraphs
- Avoid complex terminology without explanation
- Use bullet points sparingly and only when improve readability

3. prohibited language & terms of avoid these AI and marketing Buzzwords
Leverage
Utilize (use "use" instead)
Optimize
Robust
Solution
Seamless
Game-changing
Innovative
Revolutionary
Next generation
State-of-the-art
Change in thinking
Synergy/synergistic
Best-in-class 
Paradigm shift 

See files in https://github.com/bflaven/ia_usages/tree/main/ia_using_claude/ia_claude_create_agents

Two additional investigations discovered along the way and much more practical.

3. Extra-stuff: CLAUDE FOR SHEETS, using claude and prompt in Google Sheets

You can connect Claude to Google Sheets using an API key. ChatGPT offers similar connections, and works with more programs like Canvas too. The main idea stays the same though: you set up your prompts and content, then choose how you want the results to look.

My prompt to get some examples from Claude directly, which is also my user story, was as follows:

As a specialist for Claude for Sheets can you give a set of functions that helpful to automate some task on SEO with the ability to input text from a column.

Why connect Google Sheets with an API key?

It gives you the best of both worlds:

  • You get to use prompts for everything from writing content to office tasks as you learn more about them.
  • You don’t need to build special screens to use AI.
  • You can better manage how much you spend on input and output tokens.

Here are some examples of functions available in Google Sheets.

=CLAUDE("Generate 5 long-tail keyword variations for: "&A2, "claude-3-haiku-20240307", "temperature", 0.2)
=CLAUDE("Create an SEO-optimized H1 headline for this topic: "&A2&". Include the keyword: "&B2, "claude-3-haiku-20240307", "temperature", 0.5)

See files in https://github.com/bflaven/ia_usages/tree/main/ia_using_claude/ia_claude_for_sheets

4. Extra-stuff: Using IA to make a presentation using Reveal.js or PowerPoint

The idea behind these examples are to show how to create a “roadmap” for your presentation or video. The roadmap is the story that guides your slides and makes your message clear for work or personal projects. You can then easily turn this roadmap into a Reveal.js or PowerPoint presentation. This approach saves you lots of time and helps you get more done.


# Clone the git repo Reveal.js
git clone https://github.com/hakimel/reveal.js.git
cd reveal.js

# Install dependencies
npm install

# Use Reveal.js

# start local
npm start

# check localhost
http://localhost:8000

See files in https://github.com/bflaven/ia_usages/tree/main/ia_using_claude/generate_ppt_python

More infos