The Warp Agent CLI
warp.dev
warp.dev
Why is every company seemingly trying to re-implement agents themselves rather than making their product work with the existing ones?
You can't raise a $50 million series b on making a nice terminal. we already have lots of nice terminals for free
So the open source code on GitHub is the library that agents need to train on, to do all the economically useful work for 1/100th of the cost or close to free.
That is where it is going and the raised money will be going to the agents.
But I understand that's not a viable business model. I'm back on iTerm2 now cause Warp isn't approved for use at my job (and never would be with all their AI pushing), and I've managed to pound iTerm 2 into something usable enough but I wish every terminal would copy Warp's text input style.
That was also the thing that destroyed it for me, since navigating between history entries started getting more and more unreliable and sometimes clicking on the input box would randomly resize the entire window or move it between screens. I don't entirely know how they managed that one.
To be more specific about what I was saying, on Mac, in any standard text field I can hold Cmd and hit left or right to jump to the start or end of a line, or hold Option to jump a word at a time, and if I hold shift it'll select text. Option-delete deletes the last word, etc.
I've managed to get iTerm2 to do a lot of this stuff but, for example, Option-shift-left, instead of highlighting the last word, spits out the text ";4D"
Bash (also works in zsh): https://www.gnu.org/software/bash/manual/html_node/Readline-...
And for the AI part, the AI command generation (from plain text) was great. Also the follow-up command suggestions were handy sometimes.
Too bad Warp entered enshittification (already some time ago). Any (open source) alternatives for the good parts?
Another great option (not mine) is Orca [3] and it's a lot more mature. Really nice to use.
It was also a bit of an experiment with a larger Rust codebase, which is outside my usual areas. I did pull out powershell since this was just for personal use on macOS. Been using it today and enjoying it in place of Ghostty, but merely an experiment and learning project:
Why does any company do anything? Money, of course.
Well, i'm trying to do the later, and it is very hard to position the product and differentiate without vertical integration. When Warp owns both agent and the terminal, they can a) give terminal for free and charge on the agent, and b) lock the user in, so it is a good business but poor user experience.
on the other hand, i have to make sure every agent is working which more surface area to maintain, and i'm also limited to the common subset of features. Also it is much harder to convince folks to switch.
Lets try: I support Wezterm (and ghostty, and others) with 40+ different agents (Claude, Codex, etc) in agentastic.dev, would you give it a try?
> please implement the fix
$ please: command not found
The main reason you might try it is the shell access. We made CLI commands runnable in an agent session in the foreground, similar to our terminal agent. It means you can `cd` into different directories during the session, or run interactive commands like vim or REPLs in the session. One reason is convenience (it's nicer to cd around projects and keep the conversation going) and another is visibility (you can run a REPL in the foreground and tag in an agent to help and control that REPL). We're curious how others like working this way. It's pretty convenient.
They hype up not needing a third party binary to remote and run agents against SSH'ed systems, but how does this "cloud agent" work then? Are they shelling you through their own back-end? Because that sounds bad and probably wrong? Maybe its a different use case.
I've read the product description and I don't see anything that's truly unique
Looks like they let you use your LLM of choice on anything your terminal can SSH into using their TUI without having to install Claude Code or their agent on a remote system. Also, you are able to have multiple shell instances as you would on tmux in one session, all with the LLM fully aware of each one (not sure if its a model per session, or multiple, or what).
It's spawning PTYs but does it support SSH'ing into a remote session and running commands from a remote shell? That's something I built into zmx in order to scratch that "debugging in a remote VM" itch.
Which, given open-source, is just going to be a dead-end, even before AI. Because software likes to be modifiable, and this sort of thing is literally at the center of people who like to modify software.
But, hey, if something good comes of it, fantastic. I mostly the think of Docker, an unqualified success in software and probably a huge failure in business.
And the latter actually makes this a double-good; a "successful" docker would have somehow monopolized some very important part of the process, and docker is too good to be gatekept like that.
I would be fine being a business “failure” by that definition.
I've disabled it all because it gets in the way.