build-langchain-web-browsing-agent-with-mcp

April 23, 2025 · View on GitHub

  • Created at: 2025-04-23
  • Created by: 🐢 Arun Godwin Patel @ Code Creations

Table of contents

Setup

System

This code repository was tested on the following computers:

  • Mac Sonoma

At the time of creation, this code was built using Python 3.13.2

Installation

  1. Install virtualenv
# 1. Open a CMD terminal
# 2. Install virtualenv globally
pip install virtualenv
  1. Create a virtual environment
python -m venv venv
  1. Activate the virtual environment
# Windows
.\venv\Scripts\activate
# Mac
source venv/bin/activate
  1. Install the required packages
pip install -r requirements.txt
  1. Run the modules
python airbnb.py
python browser.py

Walkthrough

Code Structure

The code directory structure is as follows:

build-langchain-web-browsing-agent-with-mcp
└───config
|   └──airbnb_mcp.json
|   └──browser_mcp.json
│   .env
│   .gitignore
│   airbnb.py
│   browser.py
│   package-lock.json
│   README.md
│   requirements.txt

The airbnb.py & browser.py files are the entry points of the web browsing agents. These are the modules you need to run from the command line.

The config/ folder contains JSON files that stores the server configurations for the AirBnB and web browsing MCP server.

The .env file contains the environment variables used by the application.

The .gitignore file specifies the files and directories that should be ignored by Git.

The requirements.txt file lists the Python packages required by the application.

Tech stack

AI

  • LLM: Anthopic Claude
  • Orchestration: LangChain

Tool use for Agents

  • Model Context Protocal: mcp-use

MCP servers

  • Web browsing: Playwright & Google Chrome
  • Travel: AirBnB

Build from scratch

This project was built using Python, LangChain and the Anthopic Claude LLM. A simple text prompt is provided to the AI Agent which makes use of it's available tools via MCP servers to gather external information to fulfil the query.

1. Create a virtual environment

python -m venv venv

2. Activate the virtual environment

# Windows
.\venv\Scripts\activate
# Mac
source venv/bin/activate

3. Install the required packages

pip install -r requirements.txt

4. Setup the config files

The config/ folder contains JSON files that stores the server configurations for the AirBnB and web browsing MCP server.

First let's set up the AirBnB MCP server. Create a file called airbnb_mcp.json in the config/ folder and add the following content:

{
    "mcpServers": {
        "airbnb": {
            "command": "npx",
            "args": [
                "-y",
                "@openbnb/mcp-server-airbnb"
            ]
        }
    }
}

Next, create a file called browser_mcp.json in the config/ folder and add the following content:

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": [
        "@playwright/mcp@latest"
      ],
      "env": {
        "DISPLAY": ":1"
      }
    }
  }
}

5. Create the Python modules

Next, we create the Python modules that will be used to run the AirBnB and web browsing agents. Create a file called airbnb.py in the root directory and add the following content:

import asyncio
import os
from dotenv import load_dotenv
from langchain_anthropic import ChatAnthropic
from mcp_use import MCPAgent, MCPClient

async def main():
    # Load environment variables
    load_dotenv()

    # Create MCPClient with Airbnb configuration
    client = MCPClient.from_config_file(
        os.path.join(os.path.dirname(__file__), "config/airbnb_mcp.json")
    )

    # Create LLM - you can choose between different models
    llm = ChatAnthropic(model="claude-3-5-sonnet-20240620")

    # Create agent with the client
    agent = MCPAgent(llm=llm, client=client, max_steps=30)

    try:
        # Run a query to search for accommodations
        result = await agent.run(
            "Find me a nice place to stay in Belfast for 2 adults "
            "for a week in August. I prefer places with a balcony and "
            "good reviews. Show me the top 3 options.",
            max_steps=30,
        )
        print(f"\nRESULT:\n\n{result[0]['text']}\n\n")
    finally:
        # Ensure we clean up resources properly
        if client.sessions:
            await client.close_all_sessions()

if __name__ == "__main__":
    asyncio.run(main())

Next, create a file called browser.py in the root directory and add the following content:

import asyncio
import os
from dotenv import load_dotenv
from langchain_anthropic import ChatAnthropic
from mcp_use import MCPAgent, MCPClient

async def main():
    # Load environment variables
    load_dotenv()

    # Create MCPClient from config file
    client = MCPClient.from_config_file(
        os.path.join(os.path.dirname(__file__), "config/browser_mcp.json")
    )

    # Create LLM
    llm = ChatAnthropic(model="claude-3-5-sonnet-20240620")

    # Create agent with the client
    agent = MCPAgent(
        llm=llm,
        client=client, 
        max_steps=30,
        verbose=True
    )

    # Run the query
    try:
        result = await agent.run(
            "Find the best restaurant in London USING GOOGLE SEARCH",
            max_steps=30,
        )
        print(f"\nRESULT:\n\n{result[0]['text']}\n\n")
    finally:
        # Ensure we clean up resources properly
        if client.sessions:
            await client.close_all_sessions()

if __name__ == "__main__":
    asyncio.run(main())

6. Run the agents

To run the AirBnB agent, run the following command in the terminal:

python airbnb.py

To run the web browsing agent, run the following command in the terminal:

python browser.py

This completes the setup of our web browsing agents!

Happy coding! 🚀

🐢 Arun Godwin Patel @ Code Creations