N8N MCP Agents are INSANE!

Key Concepts

na10 mCP Agents: AI agents within the na10 platform that leverage Model Context Protocol (mCP) servers to interact with various platforms and services.
mCP (Model Context Protocol): A protocol that allows AI agents to access and utilize tools and data from different applications and services in a standardized way.
Self-Hosting: Hosting na10 on a private server, enabling the use of Community Nodes and mCP clients.
Community Nodes: Custom nodes that can be installed in na10 to extend its functionality, including mCP server integration.
AI Agents: Intelligent systems that can perform tasks and make decisions based on data and instructions.
Brave Search API: A free API from Brave Search that allows AI agents to perform web searches.
Gemini: Google's AI model, highlighted for its new Canvas feature for coding and audio overview capabilities.
Mistral Small 3.1: A French AI model that outperforms comparable models like Gemma 3 and GPT-4 Mini in certain benchmarks.
Open Router: A platform that provides access to various AI models, including Mistral Small 3.1.
Live Weave: An online tool for previewing HTML, CSS, and JavaScript code.
Visual Studio Code: A popular code editor that can be integrated with AI models like Mistral Small 3.1 through extensions like Klein.

Setting Up na10 mCP Agents

Overview of mCP Agents

mCP agents in na10 allow AI agents to interact with various platforms like Airbnb and Brave Search.
Example: An AI agent can search Airbnb for four-bedroom apartments in Manchester and provide a structured breakdown of listings, including prices, reviews, and available dates.
The AI agent can also perform web searches using the Brave Search mCP, providing real-time information like weather updates.

Setting Up na10 on El IO

El IO: A platform for deploying and hosting na10.
Create a new service on El IO and verify it.
Select na10 and configure the server (e.g., Airbnb Asia for SGD currency).
Choose a package (medium recommended initially).
Provide necessary information and create the service.
Approval is required, potentially involving adding a credit card.
Once deployed, access the admin details and set up admin login credentials.

Installing Community Nodes

Self-hosting is required to install Community Nodes.
Navigate to Settings > Community Nodes in na10.
Install the mCP server package using the package name: n10-nodes-mCP.
This allows the use of mCP nodes in workflows.

Creating a Workflow with mCP Clients

Create a new workflow in na10.
Add an mCP Client node and select "List Available Tools."
Add a new credential for the mCP server (e.g., Airbnb).
Airbnb Setup:
- Command: MPX
- Arguments: dy open-bmb-mCP-server --ignore-robots-txt
- Environment: Leave blank (no API key required for Airbnb).
Add another mCP Client node and select "Execute Tool."
Select the tool name from the previous "List Available Tools" step.
Hardcode the location (e.g., Manchester) or use AI to define parameters.

Setting Up Brave Search mCP

Command: MPX
Arguments: dy @model-context-protocol/server--search
Environment: Brave_API_key=<your_api_key>
Get a free API key from brave.com/search/api (2,000 free queries).

AI Agent Integration with mCP Servers

Add an AI Agent node to the workflow.
Connect it to an OpenAI connection (API key required).
Add an mCP Client node for "List Tools" and select the appropriate credentials (e.g., Airbnb).
Add another mCP Client node for "Execute Tool."
Set the tool parameters to be defined automatically by the model (AI).
Use the expression to define the tool parameters.

Advantages of mCP Agents

Efficiency: Requires fewer tools compared to traditional AI agent setups.
Simplicity: Easier to set up and configure.
Power: Provides more freedom and control over AI agent behavior.
Real-Time Data: Access to real-time information through web searches.
Customization: Ability to customize search results and filter information.

Gemini AI Model Updates

New Features

Canvas: An interactive space for refined documents and code.
Audio Overview: Turns files into podcast-style discussions.
Deep Research: Creates in-depth reports from web searches.

Using Gemini Canvas

Go to g.o/app and select Canvas.
Use prompts to generate code or designs (e.g., "Create a landing page for an SEO agency").
Share the canvas via a subdomain.
Preview and edit code directly in the canvas.

Using Gemini Audio Overview

Upload a file (e.g., a Google Doc).
Select the option to generate an audio overview.
The AI will create a podcast-style discussion based on the document.
Share the audio overview via a Gemini subdomain.

Comparison with Other AI Models

Gemini's new Canvas feature produces better front-end designs compared to ChatGPT and Claude.
Gemini is faster at coding and generating outputs compared to ChatGPT.

Mistral Small 3.1 AI Model

Overview

A French AI model that outperforms comparable models like Gemma 3 and GPT-4 Mini in certain benchmarks.
Open-sourced and available on Open Router.
Multimodal and multilingual.
Long context window (128k tokens).

Using Mistral Small 3.1

Access it through Open Router.
Use it for coding, reasoning, dialogue, and visual understanding.
Download it locally using O llama.
Integrate it with Visual Studio Code through extensions like Klein.

Performance

Fast response times.
Low cost (less than half a cent per use on Open Router).
Good for coding, but not as strong as Gemini for front-end design.

Conclusion

The video provides a detailed guide on setting up and utilizing na10 mCP agents, highlighting their efficiency and power in interacting with various platforms. It also showcases the new features of Google's Gemini AI model, particularly its Canvas and audio overview capabilities, and introduces Mistral Small 3.1 as a competitive open-source AI model. The presenter emphasizes the importance of adapting to these AI advancements to stay ahead in the SEO and business landscape.