Ollama MCP AI Agents: Build POWERFUL & FREE Local AI in Minutes

Key Concepts

• Local AI Processing: Running AI models directly on your device, ensuring data privacy and eliminating reliance on cloud services.
• Ollama: A tool for running large language models (LLMs) locally.
• mCP (Model Context Protocol): A universal AI tool standard that allows AI agents to access and utilize various tools.
• AI Agents: Software entities designed to perform specific tasks autonomously, often leveraging LLMs.
• Prais Agents: A Python library simplifying the creation of AI agents with LLMs and mCP tools.
• Gradio: A Python library for creating user interfaces for machine learning models.

Local AI Agents with mCP Tools using Ollama

Introduction

The video demonstrates how to create AI agents that run 100% locally using Ollama and leverage mCP tools for enhanced functionality. This approach offers privacy, cost-effectiveness, and unlimited usage compared to cloud-based AI services.

Benefits of Local AI Processing

• Privacy: Data never leaves your device.
• Cost: No monthly subscription fees.
• Unlimited Usage: No API rate limits.

Ollama Setup

1. Download Ollama: Go to https://ollama.com/ and download the appropriate version for your operating system.
2. Pull the LLM: Open your terminal and run ollama pull llama3:latest to download the Llama 3 model.

Creating an AI Agent with Prais Agents and mCP

1. Install Required Libraries: Run pip install prais agents gradio in your terminal.

2. Create app.py: Create a new Python file named app.py and open it in your code editor.

3. Import Libraries: Add the following lines to app.py:

   from praison_agents import agent, mcp
   import gradio as gr
   

4. Define the Agent: Create an AI agent using the agent function, specifying the model and instructions:

   def search_airbnb(query):
       search_agent = agent(
           instruction="You help book apartments on Airbnb",
           llm="llama3:latest",
           tools=["airbnb"]
       )
       result = search_agent.start(query)
       return result
   

   • instruction: Defines the agent's purpose.
   • llm: Specifies the LLM to use (in this case, Llama 3).
   • tools: A list of mCP tools the agent can access.

Adding mCP Tools

• mCP Servers: The video mentions a list of available mCP servers, including Airbnb.
• Adding a Tool: Each mCP server has a specific command to add it to the agent. For example, the Airbnb tool requires running a command like npx -y @modelcontext/cli add airbnb --ignore-robots-txt.
• Environment Variables: Some mCP tools require API keys or tokens, which can be set as environment variables (e.g., BRAVE_API_KEY for the Brave Search tool).

Creating a User Interface with Gradio

1. Define the Interface: Use Gradio to create a user interface for interacting with the agent:

   iface = gr.Interface(
       fn=search_airbnb,
       inputs=gr.Textbox(lines=2, placeholder="Search for apartments in Paris for two nights 5th and 6th April 2025"),
       outputs="text",
       title="Airbnb Booking Assistant",
       description="Enter your search query to find apartments on Airbnb."
   )
   iface.launch()
   

   • fn: The function to call when the user submits the query (in this case, search_airbnb).
   • inputs: Defines the input field (a text box).
   • outputs: Specifies the output type (text).
   • title and description: Set the title and description of the user interface.

2. Launch the Interface: Run demo.launch() to start the Gradio interface.

Running the Application

1. Run the Script: Open your terminal and run python app.py.
2. Access the Interface: Open the URL provided in the terminal (usually http://127.0.0.1:7860/) in your web browser.
3. Enter a Query: Type your search query in the text box and click "Submit."
4. View the Results: The AI agent will process the query using the Airbnb mCP tool and display the results in the output area.

Example Query

• "Paris for two nights 5th and 6th April 2025"

Additional Resources

• mCP Documentation: Clear documentation is available for mCP.
• Example mCP Servers: Examples are provided for Brave Search, GitHub, and Google Maps.

Conclusion

The video demonstrates a simple and powerful way to create local AI agents with access to various tools using Ollama, Prais Agents, and mCP. This approach offers privacy, cost-effectiveness, and flexibility, enabling users to build custom AI solutions tailored to their specific needs. The use of Gradio further simplifies the process by providing an easy-to-use interface for interacting with the agents.