It made me cancel my Cursor subscription. Unbelievable things are happening in the open source community.
Your tool for real work.But it's an agent.
Turn your tokens into results that matter. Research across ten thousand sources. A deck people actually get. Company work, experiments, backends, frontends, code. Whatever you need done.
One command, one tool
bun install -g omo-ai@beta$ mass ulw 12k sources → ranking model → deck
Workflow · market research, 12,000 sources
Wave 1 · 0/1 done
Wave 2 · 0/2 done
Wave 3 · 0/3 done
Wave 4 · 0/3 done
Wave 5 · 0/1 done
- 69k GitHub Stars
- 3.8M+ Total Downloads
- 224.4k Monthly Downloads
We'd rather you never had to read what's below.
We use this era's technology for one purpose only: helping you get your work done. You don't need to study the tech. Focus on your work, and the more you use it, the more it becomes the tool that fits you.
Still curious? This is for you.
ultrawork
Skip the study. If a task feels hard, add the word ultrawork to your prompt. The agent decides how to attack it and drives to the best result on its own.
›make the quarterly report actually readable ultrawork
- reads the whole codebase before touching a line
- plans, then proves each step before moving on
- verifies the result on the real surface, then stops
Mix models without thinking about it.
Use the frontier models you love together with the cheap ones. Just sign in. Every model ships with a system prompt tuned for it, so you never pick the wrong one.
Tuned profiles shipped with OmO
Hundreds of agents working for you. One keyword.
Type mass ulw and the whole job becomes a graph of agents: deep research over thousands of sources, a full machine-learning pipeline, a migration nobody wants to do. Each part runs on the model that fits it, in parallel, and gets checked before anything is called done.
market research, 12,000 sources
wave 1/5 · 0/10 done · 7 models
Wave 1 · 0/1 done
Wave 2 · 0/2 done
Wave 3 · 0/3 done
Wave 4 · 0/3 done
Wave 5 · 0/1 done
Kibitzer. Your agent doesn't forget you, or your work.
A small, cheap model sits beside the expensive main agent, reads what you've done before, and nudges the main agent before it repeats a mistake. The expensive model stays sharp; you stop re-explaining yourself.
High-quality skills. Only the ones you need.
Your model is smart, but it still doesn't know a lot about solving real-world problems. OmO ships the missing knowledge as skills and loads only what the task needs.
frontenddesign system, layout, motion, visual QAdebugginghypothesis loop, failing test firstulw-researchthousands of sources, claim graph, cited reportdata-scientistDuckDB, Polars, charts from raw filesgit-masteratomic commits, rebase, bisectultimate-browsingJS pages, logins, screenshotsimagegenphotos, mockups, transparent assetsvisual-qa375 / 768 / 1280 screenshot gateast-grepstructural search and codemodsreview-workone reviewer, real QA evidenceprompt-engineeringsmallest correct edit to a promptlsp-setuplanguage servers wired for the agent
And everything else you'll touch is crafted.
You won't see most of this. That's the point.
Code-mode tool calls
Independent tool calls run together in one small program instead of one slow round trip each.
Tool-call absorption
Dozens of read, search and lookup calls collapse into a single step. Fewer turns, less waiting, fewer tokens.
Tool-call correction
Malformed arguments are repaired before they execute, so a typo never becomes a wasted turn.
Team mode
A lead agent runs a team of specialists in parallel when a job is big enough to split.
Monitors, not polling
Waiting on a build, a deploy or a file is a subscription. The agent keeps working and wakes when it lands.
Goals that survive
Long work is checkpointed with evidence, so a session restart resumes where it stopped.
Hot-reloading everything
Config, skills and prompts reload live. No restarts to try a change.
And from your SlackDiscordTeamsWebexTelegramWhatsAppLINEWeChatiMessageInstagramKakaoTalkChannel TalkLark
Talk to your agent where you already are. It messages as you, not as a bot, on every platform you use.
*Coming soon. Messaging platforms are being wired in one by one.
What people say.
If Claude Code does in 7 days what a human does in 3 months, OmO does it in 1 hour.
You guys should pull this into core and recruit him. Seriously. It's really, really, really good.
Install
One command. Then get back to work.
Install OmO, open your project, describe the job. That's the whole setup.
bun install -g omo-ai@beta