What is Mcp Chatbot Demo

mcp-chatbot-demo is an example chatbot that utilizes agentic workflows to facilitate interactive conversations. It is built using a ReactJS web interface and integrates with various components including an agent powered by Anthropic and a PostgreSQL database for storing chat histories.

Use cases

Use cases for mcp-chatbot-demo include automating customer service inquiries, providing personalized learning experiences, facilitating user engagement on websites, and assisting in data retrieval tasks.

How to use

To use mcp-chatbot-demo, users can interact with the chatbot through the web interface. The chatbot processes user inputs and generates responses using the underlying agent, which can be enhanced by connecting to additional Secure MCP servers.

Key features

Key features of mcp-chatbot-demo include a user-friendly web interface, integration with powerful language models via Fast-agent, secure communication through Minibridge, and the ability to store conversation histories in a PostgreSQL database.

Where to use

mcp-chatbot-demo can be used in various fields such as customer support, virtual assistants, educational tools, and any application requiring interactive dialogue systems.

Chatbot example with agentic workflows

Overview

This an example chatbot with agentic workflows.

Architecture

In this demo, we will have the following components:

• A Web Interface, which presents a chat to the user. This is a simple ReactJS application making calls to an Agent
• An Agent uses Anthropic (optionnally, can use other LLMs as by supported by Fast-agent)
• A Database uses a postgres database to store and retrieve chat histories and conversations
• Some Secure MCP servers which provide additional capabilities to the Agent.

[!TIP]
Add more Secure MCP Servers as needed to make your Chatbot more powerful!

[!NOTE]
Understand the benefits of these MCP Secure Server by visiting our Github mcp-servers-registry repository

Framework used:

• Fast-agent which supports the maximum features with regards to MCP.
• Minibridge makes it secure and production ready in the secure MCP servers.
• Descope optionnally adds authentication and authorization support.

Enterprise Ready MCP servers used:

• mcp-server-fetch Dockerfile Container
• mcp-server-brave-search Dockerfile Container
• mcp-server-github Dockerfile Container
• mcp-server-sequential-thinking Dockerfile Container
• mcp-server-memory Dockerfile Container
• mcp-server-microsoft Dockerfile Container

Installation and Setup

Getting started

• Clone the repository:

  git clone <repository-url>
  

Deploying this on K8s

Follow the instructions here

Trying out the application with Docker Compose

Follow the instructions here

Developing ui and agent (using Docker compose)

Follow the instructions here

Support for TLS

[!TIP]
For ease of use, we are using tg which is a simple
TLS generator CLI: tg cert --name server --ip 127.0.0.1 --dns localhost

[!WARNING]
The certificates provided are only for ease of use and MUST not be used in production.

All certificates for MCP servers already exist in certs folder. You can update the certs/init-certs.sh

API Endpoints

• GET /health: Health check endpoint
• POST /api/v1/chat: Send a message to the chat API
• GET /api/v1/conversations: Get all conversations for the current user
• GET /api/v1/conversations/{conversation_id}/messages: Get all messages for a specific conversation

Running Tests

pytest