AISHAH SOFEA
Me
Back

The CLI Renaissance

24 April 2026

cli ai agents developer-tools

Why the Terminal Is Having a Moment

I’ve been noticing a lot more CLI activity lately: GitHub, GitLab, Stripe, Vercel, Supabase. Seems like every dev tool company is investing more in their CLI than they used to.

At first, I thought I was just paying more attention. CLIs aren’t new after all. curl, git, kubectl have been around forever. But the volume felt too high to be coincidence so I looked into it a bit.

Turns out it makes sense when you consider AI agents. Tools like Claude Code, OpenAI’s Codex CLI, and GitHub Copilot CLI are all terminal-native. They run commands, read the output, and loop until the job is done. The terminal is basically their operating environment. Claude Code alone went from zero to $1 billion in annualized revenue within six months, and while it’s since expanded into VS Code and desktop apps, the terminal is still its native habitat.

Why the Terminal Specifically

What’s interesting to me is why the terminal specifically. AI models can click buttons now, they can see screens, so why the CLI?

I think it comes down to what agents actually need. A bash command returns an exit code. Zero means it worked, non-zero means it failed. Nothing to interpret. A GUI on the other hand, doesn’t give you that. The agent has to look at pixels and decide what happened, and that’s where things start to go wrong. And when you’re running dozens of steps autonomously, interpretation compounds into mistakes.

There’s also the cost angle. A single screenshot of a GUI costs an LLM thousands of tokens to process. A shell command costs a fraction of that. Multiply that across an entire workflow and the CLI isn’t just cleaner, it’s significantly cheaper to operate.

And then there’s composability. The Unix idea of small tools piped together turns out to be exactly how agents like to work. That kind of chaining just isn’t possible when everything lives behind a UI.

So in a way, the terminal was always well-suited for this. We just didn’t have agents until now.

The MCP Land Grab

This also explains the rise of MCP (Model Context Protocol), a standard Anthropic introduced in late 2024 that lets AI agents connect to external tools and services. Most MCP servers are essentially wrappers around CLIs, communicating over stdin and stdout. Stripe, Neon, Supabase, Vercel, Netlify, Cloudflare all shipped MCP servers in the past year. Every major cloud provider basically rushed to make themselves accessible to agents, and the CLI was the fastest path to get there.

Your Customer Is the Agent

Vercel’s CEO Guillermo Rauch said something recently that stuck with me. He mentioned that 30% of the apps on Vercel now come from AI agents, not humans. And then he said this:

Your customer is no longer the developer. Your customer is the agent that the developer or non-developer is wielding.

That reframe is what makes this more than a dev tooling trend.

Because if the primary user of your product is no longer a human clicking through a dashboard, the whole product surface changes. Salesforce is already thinking about this. They announced something called “Headless 360” where they make their entire platform accessible without the UI, explicitly so agents can use it. Their co-founder asked: why should you ever log into Salesforce again?

I don’t think most companies have fully sat with that question yet. But I think they will have to.

The terminal was always there. Turns out it was just waiting for the right kind of user.

References

  1. Claude Code Revenue — Threads / GrowthLearner, “Claude Code: $2.5B Run-Rate Revenue by February 2026” (2026)
  2. Vercel CEO on AI Agents — TechCrunch, “Vercel CEO Guillermo Rauch signals IPO readiness as AI agents fuel revenue surge” (Apr 13, 2026)
  3. Introducing MCP — Anthropic, “Introducing the Model Context Protocol” (Nov 2024)
  4. Salesforce Headless 360 Announcement — Salesforce, “Introducing Salesforce Headless 360. No Browser Required.” (2026)
  5. Salesforce Headless 360 Analysis — VentureBeat, “Salesforce launches Headless 360 to turn its entire platform into infrastructure for AI agents” (2026)