Skip to Content
Introduction

YOLO Studio

All your coding agents. One secure workspace.

YOLO Studio is a cloud workspace where you direct AI coding agents. You describe the work, agents build it, and you review what comes back.

A workspace gives you a tiled desktop, an isolated container, terminals, live previews and coding agents (Claude Code, Codex, Muse Code and more) that can read your repo, write code and push branches. Agents coordinate through a shared board.

How work is organized

Work is described on a and executed in .

A is one unit of work with a prompt — the only work type there is. Cards can be blocked by other cards, and that is the only relationship between them: a dependency edge states necessity, not preference. Verification is not a special mode; a review is just a card that depends on the work it reviews.

The Background Runner dispatches eligible cards; when a card runs it gets a — a full git clone in the same container, so parallel agents share a machine but never share a git index. What the card produced is recorded separately from where anyone dragged it: admission is intent, completion is truth.

Loops iterate an agent toward exit checks you define, and Goals turn an outcome you want into a plan of steps you queue as cards and results you review.

See Concepts for the canonical vocabulary.

What a workspace gives you

  • Isolated agent : each running card gets its own git clone and , so parallel agents never share a git index.
  • Boards you and your agents both edit — intent is expected to change while work is running, so the board is mutable and realtime. The kanban tile is a view of it; closing the tile changes nothing in flight.
  • Workers and scouts — a spawn that can write is a and needs a card; a is read-only, has no lane, and just reports back.
  • Cooperating tiles: split a task across parallel lanes, add watch tiles that rerun your tests when files change, and let one card’s result start the next.
  • Drive everything from — the workspace MCP server lets agents read and change boards, spawn work, report , and inspect the workspace they’re operating in.
  • Getting Started — create your first workspace
  • Workspaces — tiles, previews, server, container runtime
  • Hosting — ship a project live on a yolo.host subdomain with yolo deploy
  • Concepts — the nomenclature reference
  • Background Runner — how cards are dispatched into agent lanes
  • Goals — plan toward an outcome and review the results
Last updated on
Overview