one terminal for every agent you run
Run several coding agents at once, each in its own worktree. See what they're doing, answer the one that's stuck, ship what worked.
Free for one project. $75 once for unlimited.
Everything the work needs, in one window
What you'd otherwise assemble from six apps and a stack of tabs.
Worktree manager
Each agent gets its own checkout. They run in parallel on one repo without overwriting each other. Review the diffs, keep what worked.
Your machine, from your phone
When a pane stops to ask something, your phone tells you. Read the scrollback and answer from anywhere, encrypted end to end.
Scrollback that survives
Full history kept on disk, compressed. Reboot and the pane opens where you left it, everything still searchable.
As many terminals as you want
Split, stack, tab, and name them. Drag to rearrange, collapse to a tray, jump to any one from the command palette.
Git without leaving
Stage by hunk, write the message with AI, commit and branch in the same window. Conflicts get a resolver that has the context.
Other machines, same window
Run panes on your own server over SSH and drive them like local ones. Your laptop stays a screen.
Fast speech to text
On-device dictation anywhere in macOS, not just inside harnx. Speak a prompt straight into whichever agent is waiting.
Screenshot and annotate
Grab any region, mark it up with arrows and text, hand it to an agent. The fastest way to say "this bit, right here".
You already run more than one thing at a time
The terminal was built for one human typing one command. The IDE, for one human reading one file. Neither was built for this.
One project free. Unlimited for $75.
Every feature works on the free plan. The $75 lifts the project limit.
Free
The whole app, pointed at one project.
- One project at a time
- Every feature, nothing held back
- Unlimited panes, splits, and agent sessions
- No account, no card, no time limit
- Swap the free slot whenever you like
Unlimited
For when one project stops being enough.
- Unlimited projects, side by side
- Everything in the free plan
- One year of new versions included
- Activate on up to three computers
- Yours to keep, no renewal
Bring your own API key, or use the Claude and Codex subscriptions you already pay for. harnx never proxies or meters your inference.
Things people ask first
Short answers on purpose.
What about my existing setup?
It stays. Your editor, shell, and scripts keep working — harnx is where you do them from.
Which agents does it support?
The ones you already use, and the next ones you try. harnx doesn't pick winners.
Is this a chat app?
No. Chat is one shape work takes; most of the others aren't conversational.
Will I have to learn a new way to work?
No. What you already do gets faster. The rest is there when you want it.
Can I use harnx for free?
Yes — every feature, on one project, with no account or time limit. Remove that project to free the slot for another. $75 unlocks unlimited projects.
Is harnx a subscription?
No. $75 once, including a year of new versions. The version you buy stays yours after that.
Do I need an account?
Not for the free tier. Buying is passwordless too — Stripe takes an email for the receipt, then the app activates itself. Three computers per license.
Do I need my own API key?
For the built-in AI, yes — a key from Anthropic, OpenAI, OpenRouter, Groq, or Google, called directly with no markup. Agents like Claude Code use whatever they already have.
Where does my code run?
On your machine, or a server you own over SSH. There is no harnx-hosted compute.
One window. Every agent.
Free for one project. No account, no card.