Solo builder shipping daily — is there a minimal path into OpenRig?

[OBG-ent]

Hi! I'm a beginner shipping two small projects to GitHub every day, working with a single AI assistant. OpenRig looks amazing, but it's clearly built for bigger setups.

Is there a recommended minimal path — like the smallest useful rig for one person? Or is this overkill until I'm running multiple agents? Just trying to figure out at what point OpenRig starts paying off.

Thanks for building this in the open 🙏

[aj0413]

So the thing is that I do think even a team of three agents would be a good starting point if you’re trying to get into fully agentic workflows

And that’s a whole skillset and learning to navigate the prompts and stuff can be hard

But I’ve been experimenting with this at work and the team size just keeps growing and growing as you add projects and teams to help coordinate the teams of agents and maybe another to help trouble shoot openrig itself

I would not recommend this for a paired programming type workflow or anything hands on as it’s too much for that

If you use a Max sub, you can let it go and will halt and restart on its own 🤷‍♂️

[mvschwarz]

Thanks for asking, and for the kind words.

The smallest useful path is the guided first run in the README: a two-agent team in your own repository, one agent that makes a change and one that checks it, using the account you already have. That's the first-project starter, or first-project-claude if you're on Claude Code only, and no second subscription is needed. The getting-started guide walks you through one reviewed change: https://github.com/mvschwarz/openrig/blob/main/docs/reference/getting-started.md

To be honest about the payoff: OpenRig earns its keep once more than one agent is working, because what it adds is the coordination around them, like keeping their work in a queue, passing it between them, and bringing them back after a restart. With one assistant and small projects, you may not need it yet. The two-agent starter is a low-effort way to find out whether a second agent checking the first one's work is worth it for you.

Thanks too, @aj0413, for sharing how it's gone at work. Starting small and adding agents as the work grows is the path we'd suggest as well.

[lab1207]

Not overkill, but start smaller than the full fleet picture. Do this:

1. Run the guided first-use path: install via the release script (docs/reference/getting-started.md, linked from the README), talk to the kernel operator, and put one of your two projects on a single seat team. That is the smallest useful rig — one seat, no pods, no fleet machinery.
2. Use it for one thing first: background/async work (a seat keeps working while you do something else) or the review gate before integration. If that saves you real time in a week, add the second project.

Honest threshold: if every session is you interactively driving one assistant start-to-finish, a rig adds ceremony without payoff. It starts paying the moment you want work to continue without you watching, or a second pair of eyes (review seat) before things land. Two small daily projects with backgroundable chunks clears that bar — you don't need multiple agents to benefit, one seat running while you do the next task is already the win.