How to Set Up MCP Server? Easy MCP Server Setup Guide
Imagine having an assistant that can read the internet for you and pull out the exact information you want. That’s what an MCP server helps you do
It’s like a link that connects AI to real tools like web scrapers, so the AI can act on fresh, real-world data. If you’ve ever wished your chatbot could just get the best when you ask for something live, setting up an MCP server is the only thing that can make it happen.
In this post, we’re gonna tell you about everything, slowly and clearly, so you can get a Decodo MCP server running with Cursor, VS Code or Claude Desktop.
We’ll also learn why each step matters, we’ll also show examples of the commands you’ll type, and also give useful tips for avoiding common problems so you feel confident.
Just in plain language and helpful details (like screenshots), we’ve tried to make the MCP server setup process easy.
What Is an MCP Server?
An MCP Server, also known as Model Context Protocol Server, is a type of server that helps AI models get better at understanding real-world situations. You can think of it as a bridge that connects AI to external tools and data sources.
This makes the AI much smarter because it can pull in live information and work with real systems. An MCP server is the part that actually makes this work.
It acts like a messenger between the AI and your tools. The server talks to the right tool, gets the result and sends it back to the AI.
You can run an MCP server locally on your own computer also. For most people, it is the best way to run the server as it is simpler and safer, since your private credentials stay on your machine.
How to SetUp MCP Server?
Before you start setting up anything, you need to check a few basic things on your computer.
Node.js version 18 or higher should be installed, because Decodo’s MCP server runs on Node.
If you don’t have Node installed yet, Go to the official Node.js site >> Download the installer for your operating system >> Run it! Most installs are straightforward and the website has clear instructions
You’ll also need a text editor or IDE ready (VS Code is a great free choice) and a terminal or command prompt you’re comfortable using.
Step1)
Now let’s get into the first step!
- First, sign up at Decodo and be ready to copy your Web Scraping API username and password; those credentials are what let the MCP server use Decodo’s scraping power on your behalf.
- Next thing is to get your Decodo API credentials.
- Create a Decodo account and either choose a subscription or start a free trial if available.
- Once you’re logged into the Decodo dashboard, find the section that lists your API credentials and copy both the username and password.
- Treat these like a secret: store them in a password manager, or put them in a secure local file that only you can access. You’ll use these values later to let your MCP server authenticate with Decodo.
If they leak, someone else could use your account.
Step2)
Next, download the Decodo MCP server code so you can run it locally. Open your terminal and use git to clone the repository with:
- This command copies the project files to a new folder on your computer called decodo-mcp-server.
If you don’t have git installed, you can either install git first or download a ZIP of the repository from GitHub and unzip it into a folder.
Step3)
After cloning or unzipping, change into that folder with cd decodo-mcp-server so your terminal commands run in the right place:
- With the code on your machine, the next step is installing its dependencies and building the project so it can run. From inside the decodo-mcp-server directory, run npm install to download the Node packages the server needs.
- This step may take a little while depending on your internet speed. After npm install finishes, run npm run build to compile the project. This creates a build folder which holds the runnable server code.
Let’s say if both the command throws an error, read the terminal output carefully! Common problems include missing Node versions or network issues while downloading packages.
Fixing those usually resolves the issue.
- Once the build is done, find the server’s main file inside the build folder.
- Typically the main file is index.js located at build/index.js. You can confirm the full path by running cd build and then pwd (on macOS/Linux) or cd build followed by echo %cd% on Windows.
- Copy that full file path somewhere handy! Your MCP client configuration will ask for it.
Knowing the exact path is important because a wrong path is a very common reason clients fail to start the MCP server.
Step4)
Now it’s time to create the configuration that tells your client how to start the MCP server.
- You’ll make a file such as mcp.json and include a small JSON block that points to Node and the index.js file and which sets environment variables for your Decodo username and password.
- The file tells the client to run the server locally with a command like node and to pass the path to index.js as an argument.
- It also exports the SCRAPER_API_USERNAME and SCRAPER_API_PASSWORD environment variables so the MCP server can authenticate with Decodo automatically.
If you plan to use VS Code and want a remote MCP, you can use Smithery to manage remote installs.
- Smithery will guide you to log in with GitHub and then it will prompt for your Decodo credentials.
- Once configured, Smithery gives you a remote connection URL that clients supporting HTTP can use to connect to the MCP server.
- It also provides a CLI command to auto-install the Decodo MCP server into a profile for VS Code: the command uses npx and requires a Smithery profile name and a key.
- Using Smithery removes some manual steps and makes managing remote setups easier, but it’s optional if you’re running everything locally.
- If you prefer manual steps, set up Cursor by opening Cursor’s settings, going to Tools & Integrations and creating a New MCP Server.
- Paste the JSON configuration you created into the mcp.json field and save.
- Restart Cursor so it reads the new config.
When Cursor launches, the MCP server status icon should go from red to green if everything is correct.
If it stays red, check that the command, args path to index.js, and the environment variables are correct. Also ensure that Node is accessible from the PATH that Cursor sees.
Sometimes apps launched from the desktop don’t use the same PATH as your terminal.
- For Claude Desktop, open Claude and go to Settings >> Developer >> Edit Config.
- Inside the configuration file, add or update the mcpServers section with the same JSON entry you used for Cursor.
If you can’t find that UI, you can edit the claude_desktop_config.json file directly:
- On macOS it is usually at ~/Library/Application Support/Claude/claude_desktop_config.json
- On Windows at %APPDATA%/Claude/claude_desktop_config.json
- On Linux at ~/.config/Claude/claude_desktop_config.json.
After saving your changes, fully quit the Claude application (not just closing its window) then re-open it so the configuration reloads.
Step5)
Once the MCP server is configured, test it right away so you know everything works. A simple test is to ask your AI to scrape a public, simple page like https://news.ycombinator.com and request “latest tech stories.”
If the server is running, the AI should call the scrape_as_markdown tool and return formatted results.
Watch the client’s logs or terminal output for helpful debugging messages: successful calls usually show the tool name and the URL requested, and failures will include an error stack that tells you what went wrong.
When testing, you may also need to accept a tool permission prompt inside your client.
Many clients ask for confirmation the first time an MCP tool tries to run. This is a safety feature to prevent accidental external calls. For the first few tests, accept each request manually so you can see what happens.
Once you trust your setup, you can enable automatic approval, but be cautious with auto-approving if the machine is shared or exposed.
What tools does Decodo MCP server provide?
Decodo’s MCP server provides several built-in tools that you’ll find handy.
- The scrape_as_markdown tool fetches a web page and returns the main content as nicely formatted Markdown.
- google_search_parsed runs Google searches and gives structured results so the AI can pick the most relevant link or snippet.
- amazon_search_parsed helps fetch product data from Amazon and returns information such as product titles, prices, and ratings in a neat format.
Learning which tool is best for which job helps the AI pick the right tool automatically.
Why should we use Decodo’s MCP Server?
Decodo’s MCP is like a smart helper that connects your AI to the web in a powerful way:
- Decodo is well-known for its fast, safe, and reliable web scraping tools and with MCP, you can plug those tools directly into your AI models. This means your AI can go online, gather real data from websites, and use it instantly in its answers or analysis.
- With Decodo’s MCP, you can scrape information from almost any website, even those that rely on JavaScript or dynamic loading. This is possible because Decodo uses a massive network of over 125 million IP addresses in 195 countries, automatically rotating them to stay undetected.
- The setup works across macOS, Windows, and Linux, and you can adjust how data is rendered or parsed depending on what you need.
- Plus, it comes with built-in error handling! If something fails, it retries automatically, manages request speeds intelligently, and gives detailed error reports so your scraping stays smooth and efficient.
Conclusion
Decodo’s MCP turns your AI into a powerful web-scraping tool. Once it’s set up, you can simply type commands like “Get the latest news” or “Fetch today’s crypto prices”, and your AI will do it automatically.
While the setup requires a few technical steps (like Node.js installation and credential setup), the result is worth it.
You’ll have a smart, secure, and reliable system that can collect real-time web data, bypassing restrictions and handling complex sites, all through a simple chat interface in Cursor, VS Code, or Claude Desktop.
// Related Articles
