You own the repo, the plan, and the merge button.
Describe your idea in a sentence. 10xs plans it, breaks it into tasks, and your machines build it, with every step checked.
Connect your project and say what you need: a fix, a new feature, an upgrade. 10xs plans it, and your machines do the work, with every change checked.
Showing: the steps for a new project.
Stage 1
One sentence. Agree the facts.
One sentence about what you want to build is enough; 10xs asks nothing else up front.
10xs writes back what it understood: who it is for, what done means, the stack, and what it will not do. Say "Sounds good" or edit a fact.
Stage 2
Argue with it, then it becomes a board.
From the facts it drafts scope, architecture and the order of work. Accept, say "not quite", or regenerate. Nothing is built until you accept.
The accepted plan is decomposed into cards with dependencies, each small enough for one agent to finish in one attempt. Nothing moves until you press Start sprint.
Stage 3
Connect with one line. Watch the cards.
Paste the connect line into a terminal on your machine. Your own agents (Claude Code and Codex) claim cards.
Each card shows its attempt, what the agent did, and a conversation you can type into while it runs. If an agent needs you, the card says why in plain English.
Stage 4
Only checked work comes back.
Every finished card is checked: tests, types, build, the house rules. When every card in the sprint is done, the work comes back as one pull request on your repo.
New feedback becomes new cards; the agents run again; you review in the morning.