The $200-a-month AI. Free.
Frontier models that work through your files, in a desktop app or your terminal. Built for people who build software or do research.
ChatGPT Pro $200/moOmnirush $0
- signed up
- 49,000+
- agent hours
- 181,000+
- tokens/day
- 14B
- on Discord
- 17,000+
Help me work out why these two assays disagree, and what I should run next.
Worked for 7m 48s
Bottom line
Trust Run A provisionally; do not use Run B for quantitative reporting.
Why they disagree
The pattern is strongly consistent with evaporation/edge effects in Run B, caused by the combination of:
- Unsealed plate during incubation/early washing
- Dry oven instead of a humidified incubator
| Check | Run A | Run B |
|---|---|---|
| Mean nonblank standard duplicate CV | 2.5% | 9.8% |
| Worst standard duplicate CV | 5.2% | 13.8% |
| Perimeter sample change, B − A | — | +0.246 OD450; all 26 higher |
On your desktop. In your terminal.
One free account for both. Each view is recreated from a real session on demo files.
Works on your files.
Point it at a folder. It reads the files, runs what it needs, and tells you what it found and what to do next.
Two plate-reader runs that disagreed. It traced the gap to an edge effect and planned the rerun.
Reconciling Assay Discrepancies and Next Steps · assay-compare
Help me work out why these two assays disagree, and what I should run next.
Worked for 7m 48s ›
Bottom line
Trust Run A provisionally; do not use Run B for quantitative reporting.
Evidence from the files:
| Check | Run A | Run B |
|---|---|---|
| Mean nonblank standard duplicate CV | 2.5% | 9.8% |
| Worst standard duplicate CV | 5.2% | 13.8% |
| Interior sample change, B − A | — | −0.005 OD450 |
| Perimeter sample change, B − A | — | +0.246 OD450; all 26 higher |
Recreated from real sessions on demo files. Model names and usage figures are left out.
Built for your field.
Software and research, not just code. Something to try in four of them:
Lab sciences
“Help me work out why these two assays disagree, and what I should run next.”
Maths and theory
“Build me a study plan for measure theory, with questions that check I really get it.”
Systems and infrastructure
“Read this lock ordering and find the path that deadlocks under load.”
Research and writing
“Turn these lab notes into a proposal outline with priorities and the open questions.”
Made with Omnirush.
Live projects and words from the community, shared in our Discord.
PulsePad
by @kaizy_dude
A gamepad tester for buttons, sticks, triggers, vibration, drift and reaction time, in the browser and as a Windows app.
“I barely had to touch the code myself for a lot of the changes …”
“Created this entire 3D model with just one prompt and some reference images…”
@rajsp_29771Made a 3D village map for a Unity game
“Hate to say but my full future startup is dependent on Omnirush now”
@1vi.z3r0In the Omnirush Discord
Phone-style portfolio
by @vipulojha
A personal portfolio laid out like a phone home screen: tap an app to see projects, skills and a bio.
The Receipt Game
by @kaizy_dude
A browser puzzle game: build a shopping receipt that satisfies a growing list of odd rules, with a speed-run mode and leaderboards.
“I've moved on to Omnirush GUI which feels miles better imo, although CLI has its own perks”
@l.0.n.3In the Omnirush Discord
Free.
Frontier models at no cost, with an allowance that refills every day. No card, no paid tier.
Comparing tools? OmniRush vs Cursor, Claude Code, Windsurf, opencode and Freebuff, or as a free Cursor alternative and a free Claude Code alternative.
| Service | Price per year | Amount |
|---|---|---|
| OmnirushState-of-the-art models | $0/yr | |
| Google AI ProGemini | $240/yr | |
| ChatGPT PlusOpenAI | $240/yr | |
| Claude ProAnthropic | $240/yr | |
| Perplexity ProPerplexity | $240/yr | |
| ChatGPT ProOpenAI | $2,400/yr | |
| Claude MaxAnthropic | $2,400/yr |
- Sign-up needs a GitHub account at least 7 days old. How access worksSign up with GitHub
- A data agreement before you start. Privacy PolicyData agreement first
- Usage figures published and dated. GrowthUsage figures in public
- Every build on GitHub, with checksums. ReleasesBuilds and checksums on GitHub
Questions.
Where can I use it?
In the desktop app for Windows (x64), macOS (Apple silicon) and Linux (x64 and arm64), or in your terminal, with the command-line tool (npm i -g omnirush@latest, then omnirush login; it needs Node.js 22.19 or newer and works on Windows, macOS and Linux). One account works in both. The docs walk you through each one.
How do I get access?
Choose Start free, create your account and connect a GitHub account that is at least 7 days old; if it qualifies, you are in right away, and a few sign-ups are held for our team to review. Then download the desktop app or install the command-line tool, and sign in there with the same Omnirush account.
Who is it for?
People who build software or do research: the tiles above run from kernel work to biology. Sign-up needs a GitHub account at least 7 days old.
Is it really free?
Yes. Every account gets an allowance of frontier-model work that refills every day, with no card and no paid tier to upgrade to.
Which models are behind it?
GPT 6 Astra, GPT 6.1 Sol and GPT 6 Sol. Every chat and task starts on GPT 6 Astra, the strongest one we offer. You can switch, but you never have to.
What are swarm agents?
Several agents on one problem at once, splitting the work and checking each other's answers. Everyone has them in the desktop app: ask for a swarm when you describe the task.
What happens to what I share?
Before you start, you accept a data agreement. We collect your chats, agent sessions, workspace snapshots, and project uploads, as that agreement describes; credential files such as .env and key files are left out of uploads, unless they were committed to git. What we collect may be used to train and evaluate AI models and may be provided to research, evaluation and model-development partners. The Privacy Policy and Terms have the details.
Where do I get help, and how can I check you?
Ask in the Discord, where announcements and support live, or write to [email protected]. Our usage figures are published and dated on the Growth page, and every desktop build is on GitHub with its checksums.
Desktop app
Windows
x64
macOS
Apple silicon
Linux
x64
Debian/Ubuntu, from a terminal
curl -fsSLo omnirush.deb https://omnirush.ai/download/linux-deb && sudo apt install ./omnirush.debLinux
arm64
Debian/Ubuntu, from a terminal
curl -fsSLo omnirush.deb https://omnirush.ai/download/linux-arm64-deb && sudo apt install ./omnirush.deb
Or in your terminal
npm i -g omnirush@latestOn Windows, macOS or Linux, with Node.js 22.19 or newer. Then omnirush login, and omnirush in any project.