Warp Terminal Review: Is It Worth Replacing iTerm2?

Why Is the Terminal Still Such Sacred Ground for Mac Developers?

I've been using iTerm2 since 2014. I had profiles for every project, custom key mappings I'd half-forgotten I'd set, a colour scheme I'd tweaked over three separate MacBooks. The terminal is one of those tools developers treat like a favourite mug — functional, familiar, and weirdly personal. So when a colleague dropped a Warp Terminal invite link in our Slack channel about eight months ago, I almost didn't bother. I was wrong to hesitate. I've been running Warp as my primary terminal ever since, and it's changed how I think about what a terminal is actually allowed to be.

What Even Is the Block Model, and Why Does It Matter?

The single idea that separates Warp from every other terminal — including iTerm2, Kitty, WezTerm, and the humble Terminal.app — is Blocks. Every command you run and the output it produces get wrapped into a discrete, selectable unit. The block sits in the scrollback, visually bounded, copyable as a single action, linkable, and sharable.

This sounds small. It isn't. When I'm debugging a Docker Compose stack and I need the output from three commands ago, I don't scroll and squint — I click the block and copy it. When I'm pairing with someone remotely and I want to share exactly what my npm install spat out, I don't paste a wall of terminal noise into Slack — I share a permalink to the block. The output is no longer ephemeral noise; it's structured output you can actually work with.

The input editor at the bottom of the screen deserves its own mention. It behaves like a proper multi-line text editor — real cursor movement, select-all, undo, syntax highlighting for shell syntax. If you've ever accidentally cleared a complex command you were halfway through, you'll understand immediately why this is better. Warp is built in Rust and renders via Metal, which means the input editor and scrollback are both fast in a way that Electron-based terminals like the late Hyper simply aren't.

How Good Is Warp's AI Assistant in Real Daily Use?

Warp's AI sits inside the terminal itself — you trigger it with a hash symbol, type what you want in plain English, and it generates the shell command. I've used it to construct find incantations I'd normally have to look up, debug a gnarly ffmpeg pipe, and explain what an unfamiliar error from a Rust build log actually means.

The honest assessment: it works about as well as asking GPT-4 in a separate browser tab, but with the massive advantage that it has context about your current directory, recent command history, and shell environment. That context matters. When I typed "find all .env files modified in the last 24 hours but ignore node_modules", it produced exactly the right command first time because it knew my shell was zsh on macOS. Without that context, AI suggestions go generic fast.

Where it falls short is on privacy. Every AI query in Warp goes through their servers. If you're working inside a client's infrastructure or running commands against sensitive systems, you need to be conscious of what you're feeding the AI. Warp does offer an option to limit telemetry, but I'd be cautious in regulated environments. This is a real limitation, not a theoretical one.

What Are Warp Drive and Shared Runbooks Actually Worth?

Warp Drive is the collaboration layer — a cloud-synced library of commands, notebooks (think runbooks), and environment configurations that your team can share. I set up a runbook for our deploy sequence: tagged commands, ordered steps, annotations for the less-obvious flags. Anyone on the team can pull it up inside their own Warp instance without opening Confluence or Notion.

For solo developers, Warp Drive is a nice personal command library but not a reason to switch on its own. For teams of three or more who live in the terminal, it's genuinely compelling. The free tier is generous — you get shared notebooks and basic Drive access — but the paid Teams tier is where it becomes a proper collaboration tool. That said, the runbooks are only as good as the discipline your team brings to keeping them current, which is a human problem Warp can't solve.

Where Does iTerm2 Still Win?

I want to be honest here, because iTerm2 is a remarkable piece of software that has been free, actively maintained, and deeply loved for over a decade. It still beats Warp in several areas.

• Tmux integration is exceptional in iTerm2 — native tmux mode renders panes as real windows. Warp's tmux support is functional but not seamless in the same way.
• Profile granularity — iTerm2's profile system, with per-profile colours, fonts, key mappings, and badge text, is more mature than anything Warp currently offers.
• No account required. You open iTerm2 and it works. Warp asks you to create an account on first launch, which is a genuine friction point and a philosophical choice some developers will reject outright.
• Shell customisation compatibility — if you have an elaborate Starship prompt or a heavily customised Powerlevel10k setup, Warp will partially fight you. It renders its own input editor over your prompt area, which can create visual weirdness until you configure Warp to handle the prompt correctly.

Who Is Warp Absolutely Not For?

I mean this genuinely, not as a hedge: Warp is not the right terminal for everyone, and knowing that upfront will save you the frustration of a bad-fit install.

Privacy-first developers should look elsewhere. Requiring an account to use a local terminal application is a choice that reflects Warp's VC-backed, cloud-feature-first business model. If that bothers you architecturally, use Kitty or WezTerm — both are excellent, open-source, and completely local.

Heavy terminal-mode application users will sometimes notice friction. Tools like Vim, Neovim, Emacs in terminal mode, htop, and lazygit all use raw terminal input modes that interact with Warp's input editor in ways that occasionally feel off. Most things work, but if a significant portion of your terminal time is spent inside full-screen terminal apps, the Block model stops being relevant and the edge cases start mattering more.

Minimalists who want speed above all else and zero cloud features should try Alacritty or Kitty. They're faster to cold-start, use zero memory on features you don't need, and stay entirely off the network.

After Eight Months of Daily Use, Is Warp Actually Worth It?

Yes — with a clear-eyed understanding of what you're trading.

Warp is the first terminal I've used that felt like it was built for how I actually work in 2026, not for how Unix terminals have always worked. The Block model alone has changed my debugging workflow in ways I didn't anticipate. The AI assistant saves me genuine time, not just the novelty kind. The shared runbooks have cut onboarding friction for new team members in ways that no README ever managed.

What you're trading is: mandatory account, privacy considerations for AI features, some edge-case compatibility with heavily customised shell setups, and the long-term uncertainty of a VC-funded company whose free tier could change. Those are real costs. For me, they're worth it. For someone running security-sensitive infrastructure or deeply attached to a decade of iTerm2 muscle memory, they might not be.

If you've never felt like the terminal was holding you back, Warp won't feel transformative. But if you've ever wished your terminal output was easier to read, share, or explain — Warp is the first tool that takes that problem seriously.

Start with the free tier. Spend a week with it on a project where you're actively debugging something. The Block model will either click immediately or feel like friction — and that reaction will tell you everything you need to know.