Hey everyone!
The Google Cloud team is launching Google Cloud Live, a no fluff weekly livestream series designed around hands-on building, vibe coding, and live debugging. This livestream series will occur every Tuesday at 9A PT / 12P ET, with new developer topics weekly.
And the best part? They specifically want to hear from the DEV community and answer YOUR questions!
For their first-ever episode, hosts @stephr_wong and @greggyb, along with special guest Denise Kwan, are skipping the console and showing you how to manage your entire AI workflow from the command line.
In "Getting Started with Gemini CLI", they'll dive into:
- Setting up your terminal for AI-first development
- Managing prompts and deployments without leaving your session
- Real-time debugging of CLI extensions
- Your questions!
Whether you're curious about specific CLI features, wondering about integration with your current workflow, or have questions about AI development tooling in general, drop your questions in the comments below and the team will do their best to respond.
Your questions will help shape the conversation and make this session as valuable as possible for the community. Please note that due to time restrictions, not all questions may be answered - feel free to leave reactions to your favorite questions to help prioritize.
SO, what do you want to know about Gemini CLI? Drop them below! 👇
Top comments (10)
How does Gemini CLI handle context window management for longer debugging sessions? Specifically curious if there's a way to pin certain files or project context so it doesn't lose track mid-session when you're deep in a multi-file issue.
@ofri-peretz not sure if you caught the stream live, but you did get a response to your question @ 54:46!
I couldn't join the live session @jess , but the new context management in Gemini CLI looks huge. The ability to 'pin to memory' is exactly what’s been missing for long debugging sessions. I'll definitely be using it!
I have a good experience with gemini CLI.
How can I real avoide to fall into endless loop of working?
My solution is push that request into prompt, but on first week a time to time I fall into that trap and reach my daily limit.
I using gemini to modernize or crapy old legacy codebase written by worst CV driven philosophy. So this is a real anoying process. But definetly can help to solve.
I need my CLI agent to be able to access my Cognee MCP server. I know how to set this up in Antigravity, but can I do it with the command line tool?
What are the security principles behind building out a CLI like this?
I tried Gemini CLI a few months ago—with Gemini Pro 2.5—and it was a really interesting experience. If I can learn to use it better, I'm in!
Gemini CLI? This is interesting. It sounds stupid, but... First I learned about Copilot CLI, then I learned about Gemini CLI, and then it turns out there's ChatGPT CLI, DeepSeek CLI, Claude CLI. I feel like I'm getting old and falling behind the times! I'm definitely a couple of years behind!
Can't wait to hear from you all on your questions for our lovely hosts Steph and Greg!
As someone who's been using Gemini for coding assistance in VS Code, I'm really excited to see it move to the CLI.
There's something about the terminal that just feels more... powerful? Like, GUI is for consumers, CLI is for builders. 😎
My question for the Google team:
When working with Gemini CLI, how do you handle context windows in large codebases? If I'm in a monorepo with 100+ files, does it understand the entire project structure or just the current directory?
Asking because this is literally the only thing stopping me from fully committing to AI-assisted local development. 🙏
@googledevs @devteam