If you’ve been anywhere near the AI coding space in the past year, you’ve probably bumped into [Cline](https://github.com/cline/cline). The open-source VS Code extension has quietly built up a massive following — over 5 million developers, 57K+ GitHub stars, and the title of fastest-growing AI open-source project by contributors on GitHub Octoverse. So when the team announced [Cline CLI 2.0](https://cline.bot/cli) on February 12th, it caught a lot of attention fast. It hit [Product Hunt](https://www.producthunt.com/products/cline-4) with 250+ upvotes the day after launch, and multiple tech outlets picked up the story.

Here’s the thing — this isn’t just the VS Code extension ported to the terminal. They rebuilt the whole experience from scratch. You still get the same agentic loop (plan, reason, execute, iterate), but now it lives natively in your shell. The interactive TUI feels surprisingly polished: you can toggle between Plan and Act modes with the Tab key, watch the agent reason through problems in real-time with streamed markdown, and even switch models mid-session depending on whether you need raw power or speed.

The parallel agents feature is what really got me interested. You can spin up multiple fully isolated Cline instances — each with its own state, conversation history, and model config — and run them side by side in tmux. Imagine having one agent refactoring your auth module while another writes tests for your API layer. That’s the kind of workflow this enables.

But the real killer feature for teams might be headless mode. Pass the `-y` flag and Cline runs autonomously with full stdin/stdout support, which means you can pipe it directly into your CI/CD pipeline. GitHub Actions, GitLab CI, Jenkins — it all works. They even support JSON output for programmatic integration. This is where AI coding tools start to feel less like fancy autocomplete and more like actual infrastructure.

And if you’re not a VS Code person, ACP (Agent Client Protocol) support means you can use Cline with JetBrains, Zed, Neovim, or Emacs. So editor lock-in is no longer an excuse.

On the model side, they’re offering [Kimi K2.5 and MiniMax M2.5 for free](https://cline.bot/blog/announcing-cline-cli-2-0) for a limited time — no API key required. You can also bring your own keys for Anthropic, OpenAI, Google, Mistral, or run local models through Ollama. Installation is dead simple: `npm install -g cline`, then `cline auth`, and you’re off.

The whole thing is Apache 2.0 licensed, which matters. You can read every line of code on [GitHub](https://github.com/cline/cline), fork it, extend it, whatever you want. For an AI tool that potentially touches all your source code, that transparency counts for a lot.

Whether you’re a solo dev who lives in the terminal or part of a team looking to automate repetitive coding tasks in your pipelines, [Cline CLI 2.0](https://cline.bot/cli) is worth a serious look. It feels like a meaningful step toward AI coding agents that actually fit into real engineering workflows, not just demo videos.