Popular MCP use cases

MCP enables AI agents to interact with virtually any system or service. Here are some popular ways you can use MCP today.

MCP for software development

You know how, some days, you spend more time switching between GitHub issues, pull requests, and documentation than actually writing code? The GitHub MCP Server  can help you delegate these tasks to your AI coding assistant.

Managing pull requests using the GitHub MCP Server

For example, here we use the GitHub MCP Server to check the status of a recent pull request:

Get the status of my Context7 PR.

Here, we see Claude searching for the pull request and responding with the status.

Deleting a local branch using Desktop Commander

Next, let’s use the Desktop Commander  MCP server to delete the local branch associated with that pull request and do some housekeeping:

Delete the local branch for this PR in ~/projects/personal/mcp-servers/context7 and rename local master to main.

Here, we see Claude deleting the local branch and renaming the master branch main. How’s that for convenience?

Sure, you could have done gh pr status and git branch -d context7 yourself, but with MCP, your AI assistant could decide to do this for you, or even suggest it proactively based on your recent activity. This is just one example of how MCP can streamline your software development workflow.

Finding documentation using Context7

Context7 is one of the most popular MCP servers for software development. It can help your AI coding assistant find the most recent documentation and code examples for more than 21,000 libraries and frameworks. This helps fill in the knowledge gaps that LLMs have due to training cutoffs. Context7 can even help your assistant find code snippets for specific, older versions of libraries. This greatly reduces hallucinations.

Let’s see how it works in practice. We’ll ask Claude for help developing an MCP server:

Show me how to elicit user input using the MCP TypeScript SDK. Use Context7.

Claude uses the Context7 MCP server to find the latest documentation and code samples for the MCP Specification. It then generates TypeScript code samples for an MCP server and an MCP client.

MCP for number crunching

Despite recent advances, we all know LLMs still struggle with math and calculations. Especially when working with large numbers, the results can be wildly inaccurate.

Using MCP Run Python for accurate calculations

With MCP Run Python , we can offload any calculations to a Python environment.

Let’s ask Claude to help us with a tricky math problem:

I want to see the first moon landing live on TV. If I leave tomorrow at 6am PST, calculate the exact number of seconds I need to go back to see the touchdown three minutes after I arrive. (I need this for my time machine, obviously.) Calculate the number manually, then do the calculation again using Python. Compare the results.

Claude first tries the calculation manually, then uses the MCP Run Python server to do the calculation again. The results show that the manual calculation was off by about 53.2 days! Imagine traveling 56 years back in time, only to find out you need to wait two months to see the moon landing.

To avoid such mistakes, we recommend adding a Python server to your MCP setup, then adding the following to your system prompt or rules:

If you need to do any calculations, always use the run_python_code tool to ensure accuracy.

MCP for browser automation

Giving an LLM access to a web browser is something to behold. It can search, navigate, take screenshots, and even fill out forms.

Browser automation using Playwright MCP

We’ll use the Playwright MCP server  to automate browser tasks. This is particularly useful for testing web applications, scraping data, or even just finding information online. So let’s ask Claude to open a browser and help us with comparative shopping:

Find three cheap aluminium 60% mechanical keyboards in beige or grey on AliExpress.

Claude opens the browser, searches for the keyboards, and finds three options.

Your AI coding assistant can use the Playwright MCP server to automate testing or take screenshots of your site while it tries to fix CSS issues. The Playwright server can also be used to automate any web-based tasks, like filling out forms or scraping data.

Getting started with these use cases

You can use these examples in Claude Desktop by adding the MCP servers to your claude_desktop_config.json file:

Using MCP with different AI agents

In this guide, we focused on using MCP with Claude, but you can use MCP with any AI agent that supports MCP.

You can find a list of MCP clients and agents on the MCP website .