In Part 1, we covered how to install Gemini CLI and the basics of interacting with files using the @ symbol. Now, let’s unlock the real power of this tool.
Once you have the CLI running, you aren't limited to just chatting with text files. You can connect to databases, manage GitHub repositories, and even write directly into Google Docs.
Here are four practical ways to level up your Gemini CLI skills.
1. The Power of MCP (Model Context Protocol)
This is the most "technical" sounding part, but it is actually the most exciting.
What is MCP?
Think of the Model Context Protocol (MCP) as a "Universal USB Port" for AI.
Before MCP, if you wanted Gemini to talk to your SQL database or your Slack, you had to write complex custom code. With MCP, there is a standard way to plug these tools together.
How to Connect:
Gemini CLI acts as an MCP Client. You can run "servers" (which are just small programs that talk to your data) and Gemini can instantly interact with them.
Common MCP Servers to Try:
- GitHub Server: Lets Gemini search your repos, read issues, and create pull requests.
- PostgreSQL/SQLite Server: Lets Gemini run SQL queries to answer questions about your data.
- Filesystem Server: Gives Gemini safe access to navigate your folders (built-in).
To get started, check the /settings command or the configuration file to add an MCP server configuration.
2. Using Extensions: The Google Docs Example
Gemini CLI allows you to install extensions to expand its capabilities beyond the terminal. A great example of this is the Google Docs integration.
Why keep your AI drafts in the terminal? You can have Gemini write them directly into a document.
How to Install Extensions:
Extensions are typically installed via npm (Node Package Manager). For example, to add Google workspace capabilities, you might look for community or official packages (always check npm search gemini-cli-extension).
Real-Life Use Case:
Imagine you have finished coding a feature and need to write documentation.
- The Prompt: > "Read the code in @server.js and write a technical documentation summary. Then, create a new Google Doc named 'API Documentation' and paste the content there."
- The Result: You stay in the terminal, but the work appears instantly in your Google Drive.
3. Software Development: The GitHub Workflow
This is where Gemini CLI shines for developers. By connecting Gemini to your GitHub (via MCP or extensions), you can automate the tedious parts of coding.
Implementing Features:
Instead of switching between your browser (to read the ticket) and VS Code, do it all in one place.
- The Scenario: You have a bug reported in GitHub Issue #42.
- The Prompt: > "Read GitHub Issue #42. Then, look at @login.py and explain why this bug is happening. Finally, suggest the code changes to fix it."
Code Reviews & PRs:
You can ask Gemini to act as a senior engineer.
- The Prompt: > "I just finished writing @auth.ts. Review it for security vulnerabilities before I commit."
4. Data Analysis on the Fly
You don't need to be a Data Scientist or open Jupyter Notebooks to get quick insights. Gemini CLI can digest raw data files and give you answers.
What Data Can It Handle?
- CSV Files: Excel exports, financial data.
- JSON: API responses, configuration files.
- Log Files: Server error logs (
.txt,.log).
Example: Analyzing Sales Data
Let’s say you have a file called q1_sales.csv.
- The Prompt: > "Analyze @q1_sales.csv. Calculate the total revenue for March. Also, tell me which product category had the highest growth percentage compared to February."
Why this is powerful:
Under the hood, Gemini isn't just "guessing" the numbers. It can actually write and execute a small Python script to calculate the math precisely, then show you the result. It combines the reasoning of an LLM with the precision of a calculator.
Final Thoughts
The Gemini CLI transforms your terminal from a simple command prompt into an intelligent workspace. Whether you are fixing bugs on GitHub, analyzing CSVs, or drafting docs, the goal is the same: keep your hands on the keyboard and stay in the flow.
Have you tried connecting an MCP server yet? Let me know in the comments what you’ve built!
Credits: Concepts and structure inspired by DeepLearning.AI Course: Gemini CLI.
Top comments (0)