Google · 2026-06-05 · major
Google Ships the Colab CLI — Apache-2.0 Terminal Tool Provisions T4, L4, A100, and H100 GPUs Plus v5e1/v6e1 TPUs in Seconds and Bundles an Agent Skill for Antigravity, Claude Code, and Codex
Google released a Colab command-line tool that hands any local shell — or any coding agent — fresh Colab GPU and TPU runtimes through `colab new --gpu A100`, with `exec`, `repl`, `download`, and `log` around it.
Open-source CLI that hands local terminals — and any coding agent — Colab GPUs and TPUs on demand.
What is it?
An Apache-2.0 command-line client for Google Colab from Spencer Shumway, Tyler Pirtle, and Seth Troisi. You install it locally with `uv tool install google-colab-cli` (or pip) and then drive Colab compute from a normal shell instead of the notebook UI. It also ships a prebuilt skill file so any agent with terminal access — Antigravity, Claude Code, Codex, others — can use the CLI as a tool.
How does it work?
`colab new --gpu A100` provisions a runtime in seconds. `colab install` adds Python packages on the remote VM. `colab exec -f script.py` runs a local script there. `colab log` and `colab download` pull artifacts and replayable `.ipynb` logs back. `colab repl`/`colab console` drop into an interactive session. Supported accelerators include T4, L4, G4, A100, and H100 GPUs plus v5e1 and v6e1 TPUs; default is CPU.
Why does it matter?
Until now, hooking a Colab GPU into an agent loop meant scripting the notebook surface or wiring up the open-source Colab MCP server. The first-party CLI removes that friction: an agent on a laptop can ask for a fresh H100, run a fine-tune or eval, recover the logs, and tear down — all through the same shell tools it already uses for everything else.
Who is it for?
AI agents and developers who want one-line access to Colab GPUs/TPUs from a terminal.
Try it
uv tool install google-colab-cli && colab new --gpu T4