Google Colab CLI and Skills are out. Full Colab runtimes from your terminal.

- GPU/TPU provisioning (colab --gpu A100) - Remote script execution (colab exec) - Interactive console/REPL access - Built-in agent skill

Tell your agent "fine-tune Gemma 3 1B on this dataset" and it provisions a GPU, runs the training, downloads the adapter weights. Fully automatic.

CLI + SKILL: https://github.com/googlecolab/google-colab-cli/tree/main

Super excited about the release of the Colab CLI! 🔥

Enabling your agents to easily interact with Colab is about to unlock lots of fun experiments

https://developers.googleblog.com/introducing-the-google-colab-cli/

Learn more at https://developers.googleblog.com/introducing-the-google-colab-cli/

@osanseviero That's brilliant. Now i can debug and fine-tune SLMS even faster.

@osanseviero https://keiondra2.gumroad.com/l/KLHKLH

@osanseviero is there a proper gemini cli like codex? when i can command gemini to analyze and make a new clip from existing video?

@_philschmid great, got it featured here: https://explainx.ai/skills/googlecolab/google-colab-cli/colab-session-operator

@osanseviero colab's great for quick experiments, but real-world testing usually reveals all the flaws. hope the CLI helps bridge that gap.

@osanseviero Oh wow, finally

@osanseviero Sweet !

@osanseviero Agents spinning up Colab notebooks directly from terminal is going to unlock some crazy experiments 🔥

@osanseviero This is quite awesome. I’m really glad you did this!

@osanseviero This is amazing!!! Thank you

@osanseviero Terminal meets AI and suddenly everything starts moving faster

The Colab CLI is the missing piece for notebook-driven agent workflows — until now you'd have to awkwardly shell out or use the Colab API just to execute cells. Agent-native notebook access means an LLM can iterate on analysis in place rather than rewriting it as scripts. Research automation is going to move fast on this.

@osanseviero We had fun building it and I hope you like it! Please let us know if you have any issues or requests!

@osanseviero Colab becoming callable from agents is a bigger deal than it sounds. A lot of ML work still dies in the gap between notebook experiments and repeatable workflows.

@osanseviero Yep, the vscode tax is brutal, I just keep it terminal-first and wire agents straight into Virlo API, way less friction for quick experiments.

@osanseviero Oh wow, this one is dope! 🤩