Launching today
Operum
Your AI product team, with expert workflows built in
32 followers
Your AI product team, with expert workflows built in
32 followers
Build and launch software without managing AI agents. The Operum desktop app gives founders and developers a coordinated AI product team: PM, Architect, Engineer, Tester, Marketing, Community. Describe your goal in chat; agents plan, build in your GitHub repo, independently test changes, and help you launch. Expert workflows and human approvals are built in, with optional auto-merge. Use Cursor, Claude Code, or Codex. Engines change; your team stays. Free plan + 14-day Premium trial. No card.
Free Options
Launch Team / Built With
AWS Activate CreditsCredits that fuel your startup’s next milestone
Promoted
Hey Product Hunt!
I'm Alex, CTO and co-founder of Operum, together with @olga_primak - our co-founder and CEO.
Operum started with a problem I couldn't solve by adding more AI coding agents.
I've spent 20+ years building trading platforms and leading engineering teams. When I started building a trading app with Claude Code, I was amazed at how fast AI could write code.
But I was still doing everything around it: planning, architecture, reviewing, testing and shipping.
The AI was fast. I was the bottleneck.
I didn't need another coding assistant. I needed a team.
So we built Operum: a desktop app that gives you six specialized AI agents. Project Manager, Architect, Engineer, Tester, Marketing and Community.
It's not six agents sharing a chat. Real engineering workflows are built into the app, not just into prompts. Each agent has a defined role, it works in your own GitHub repo, the code is tested by a different agent than the one that wrote it, and you control what gets merged. Marketing and Community help you launch and support your users.
A real run, not a demo: a user filed a feature request on our public tracker. The team planned it, built it, fixed its own failing check, tested it and merged it in about 1 h 40 min. Then it drafted the reply to the user, and we approved it.
This isn't a weekend project. We build Operum with Operum: since January, our repo has merged 5,300+ pull requests, and we've shipped 100+ releases for macOS, Windows and Linux. We're on v0.62 today.
Bring the AI tools you already use: Cursor (subscription), Claude Code or Codex (API keys). OpenCode and local models are coming next. Engines change. Your team stays.
Then we discovered something bigger. Olga isn't a developer. Her background is building human teams and running businesses. Yet she used Operum to build her photography studio's website and the CRM her business runs on. So far, 8+ products have been built with Operum, by us and by early users, plus Operum itself. That's why it's for non-technical founders, developers and teams alike: you bring the idea and the decisions, and your AI team handles the development workflow.
Launch week: sign up by Sunday, Nov 16, and we'll extend your Premium trial to a full month instead of 14 days. No credit card. You'll get an email when the extra time is added. After that, stay on the Free plan or choose a paid one.
We're a two-person founding team, and we'd love your feedback:
if you build software with AI today, what's the part you still spend the most time managing yourself?
Olga and I will both be here all day answering questions.
Thanks for checking out Operum!
@alprimak Hi everyone, I'm Olga, and I'm on the Operum team with Alex
I work on our website and the launch, and I use Operum every day with my own AI team. What I like most is that I'm never wondering what the agents are doing: every change comes with a clear summary, and nothing goes out without the Tester's pass. Thank you all for checking us out today. If you try it, tell us what was confusing. That's the feedback we need most right now!
Security remains one of the biggest concerns when building software with AI. How is Operum addressing these risks and ensuring that the apps built on its platform are secure, especially compared to other vibe coding tools?
@rajat_gupta51 Great question, Rajat. No tool can promise an app is secure, so we focus on making sure AI-written code doesn't ship just because it compiles.
Maker-checker by design. The agent that writes code can't approve it. A separate Tester has to pass it, CI has to be green, and nothing overrides that. You can also require your own approval on every merge.
Agents have identities and limits. Each one has a fixed role, works in its own isolated copy of the repo, can act only on its owner's work, and every GitHub action is logged. On macOS and Linux, each agent also runs inside an OS-level sandbox. Windows support is coming.
Prompt injection is contained, not ignored. Text from issues, files or public channels is treated as data, never as commands. Only the owner can authorize privileged actions. If one agent is fooled, the checker still stands between it and production.
Secrets stay out of the chat. Keys live in your OS keychain and are never printed or committed.
Most vibe-coding tools use one agent that writes and ships in one step. Operum is a team with checks and roles. It's no replacement for a security audit on sensitive apps, but it makes "the AI quietly shipped something bad" much harder.
This product gives you not only a team of professionals but an AUTONOMOUS team. I have worked on several projects with the operum's help, it is pretty good experience compared to another tools. If you have any questions I am more than happy to help
@dmitry_stepanov37 Thanks, Dmitry, and thanks for putting Operum through real projects!
One nuance for anyone reading: the team runs on its own, but you choose how much. It can plan, build, test and merge end to end, or stop at each PR for your OK. Either way, nothing ships without the Tester's sign-off.
Curious: which project was the best fit for it, and where did you still step in yourself?
@alprimak The biggest one: I built a whole data discovery / data fabric platform with Operum. The Architect laid out the design first, the Engineer built it PR by PR, and the Tester bounced anything that broke before I saw it.
Turbo mode is what made it feel autonomous. The PM keeps pulling the next task from the backlog and assigns it without me. I mostly stepped in on product decisions and reviewing what mattered.