01 / COMPRESSION
Every request, compressed
A local optimization layer strips redundant context before it reaches the model — 60–90% fewer tokens on everyday coding, answer unchanged.
FOR DEVELOPERS WHO'D RATHER NOT PAY TWICE
Half the spend.
All of the model.
Compression compounds. On a normal day of coding that's the difference between hitting your limit at noon and it lasting until tomorrow — with the answer unchanged.
Download for macOS
free while in beta
v0.28.0 · 38 MB · requires macOS 14+ · Apple silicon
AND IT COMPOUNDS
- tokens back on one turn
- ~7,340
- further your plan goes in a day
- ~2×
- added API spend — it's your plan
- $0
Illustrative figures. Actual savings depend on your prompts and project — most coding work lands in the 60–90% range.
- ✓Works with your ChatGPT login
- ✓No API keys required
- ✓Runs locally on your machine
- ✓60–90% fewer tokens
HOW IT WORKS
Nothing new to learn.
Half the bill.
Halv sits quietly between you and the model. You work exactly as you do today — it just makes every request cheaper on the way through.
- 01
Sign in with ChatGPT
One click, the official sign-in flow. No API keys to paste, nothing to configure.
- 02
Work like you already do
Chat, or open a project and hand Halv a task. Same models, same quality you expect.
- 03
Spend half the tokens
Halv compresses every exchange in the background. Same answers, and the savings add up.
WHAT YOU GET
Built to respect your budget
— and your machine.
02 / YOUR PLAN
Use the plan you pay for
Sign in with ChatGPT and spend your own subscription limits. No metered API bills, no second invoice. An API key works too, if you prefer.
03 / IN YOUR REPO
Works in your codebase
Open a project and Halv can read files, run commands, and propose edits. Every shell command runs in a sandbox behind an approval you control.
04 / PROJECTS
On your actual codebase
Add a local folder or clone straight from GitHub. Halv binds each conversation to a working directory and learns the repo from its AGENTS.md.
05 / SAFETY
Sandboxed and local
Commands run in an OS sandbox that blocks writes outside the workspace and denies network. Halv branches your repo first, so every change is reversible.
06 / SAVINGS
See what you save, live
A running meter shows tokens and headroom saved per message and over time — so the value isn't a promise, it's a number on screen.
COMPRESSION COMPOUNDS
Same work.
Half the tokens. Every
request.
Every turn Halv trims is headroom you get back. Over a day of coding it's the difference between running out and keeping going.
standard~12,480
½
through halv~5,140
1,204
developers in the beta —
38M tokens halved this week
"I stopped watching the usage bar. It just stopped mattering."
"Same answers as raw Codex on our repo. Half the burn. We kept it."
PRICING
Start free.
Pricing lands with 1.0.
Halv is free during the beta — you already pay OpenAI, and Halv just makes that go further. Early-access users get first say and the best launch price.
Solo
Freewhile in beta
Everything you need to halve your token spend.
- Sign in with your ChatGPT plan
- Compression on every request
- Project-aware coding in your repo
- Sandboxed shell & edits
- Live savings meter
Pro
PLANNEDTBDwith 1.0
For developers who live in it all day.
- Everything in Solo
- Higher compression tiers
- Priority model routing
- Advanced approval policies
- Usage analytics & exports
Team
PLANNEDTBDlater
Shared savings across your whole team.
- Everything in Pro
- Central billing & seats
- Shared project settings
- Org-wide usage reporting
- SSO & priority support
Do I need an OpenAI API key?
No. Sign in with your existing ChatGPT account and Halv uses the plan you already pay for. If you'd rather use an API key, that works too.
How can it use fewer tokens and still give the same answer?
Halv strips what the model was going to ignore — duplicated file context, stale history, boilerplate instructions — and keeps everything that changes the answer. When a request can't be compressed safely, it passes through untouched.
Is my code safe?
Everything runs locally. Halv reads your repo on your machine, and nothing leaves except the model call itself — the same call your client already makes. Shell commands run in an OS sandbox behind approvals you control.
Which platforms are supported?
macOS on Apple silicon today, free while in beta. Windows and Linux are next — leave your email and you'll hear the moment they're ready.
How is this different from the ChatGPT app or Codex?
Same models underneath. Halv adds the compression layer, the project sandbox, and a live savings meter — you keep your workflow and your plan, and spend about half as much of it.
Get twice the mileage
out of every prompt.
Download the macOS beta, or leave your email and we'll tell you the moment Windows and Linux are ready.
v0.28.0 · 38 MB · requires macOS 14+ · Apple silicon
No spam. One email when there's news.