h0x91b/dev-3.0

Mission control for the One Person Studio — run a fleet of AI coding agents in parallel without losing your mind. Kanban + git worktrees + tmux for Claude Code, Codex, Gemini CLI, OpenCode and any shell agent. Not an IDE.

What it solves

dev-3.0 addresses the cognitive load and technical friction of managing multiple AI coding agents simultaneously. Instead of switching between editor tabs or risking file conflicts when running several agents, it provides a centralized "cockpit" where each task is isolated in its own environment, allowing a single developer to command a fleet of agents without drowning in context switching.

How it works

The system organizes work as a Kanban board where every card represents a specific task. For each task, dev-3.0 automatically creates a dedicated sandbox consisting of a git worktree, a tmux session, and a separate branch. This ensures that agents working on different tasks never collide.

Users can launch multiple agents (such as Claude Code, Codex, or Gemini) on a single task to compare different approaches or have one agent write and another review. The platform includes a built-in diff viewer for reviewing changes and a CLI that agents use to communicate status, request attention, or share artifacts (like interactive HTML reports) back to the board.

Who it’s for

It is designed for solo developers who want to maximize their productivity by running many AI agents in parallel across various projects, rather than those looking for a team-wide project management platform or a traditional AI-integrated IDE.

Highlights

  • Isolated Sandboxes: Every task gets its own git worktree and terminal, preventing file conflicts between concurrent agents.
  • Live Terminal Previews: Hover over a Kanban card to see the agent's live terminal output without leaving the board.
  • Multi-Agent Variants: Run several different agents or multiple versions of the same task on separate branches to compare results.
  • Integrated Review Workflow: A built-in diff viewer allows users to comment on code and send those reviews back to the agent as prompts.
  • Remote Access: The dev3 remote command serves the UI via browser, allowing monitoring and control from a phone or remote machine.
  • Agent-to-Board Communication: A dedicated CLI allows agents to trigger notifications, upload images, or create interactive artifacts.

Related

  • Project
    finna/Finn-loop

    A set of Claude Code skills that automate the pipeline from idea to PR using Linear and GitHub, while keeping humans as the final merge authority.

  • Project
    dedene/zentty

    Zentty is a native macOS terminal built on Ghostty that provides a tiling‑style, keyboard‑first workspace (worklanes) and a sidebar showing the status of AI coding agents (Claude, Gemini, Copilot CLI, etc.). It restores sessions, integrates with macOS (menu‑bar status, keeps the Mac awake), supports drag‑and‑drop into SSH panes, and offers a scriptable CLI. Install via Homebrew or DMG; source builds require Xcode and the GhosttyKit framework.

  • Project
    powerfooI/roamgate

    Roamgate is a browser and PWA client for the Herdr coding agent platform that enables developers to manage their workspaces, terminals, and AI agent sessions across desktop and mobile devices.

  • Project
    gendigitalinc/sage

    Sage is a security plug‑in for AI coding assistants (Claude Code, Cursor, VS Code, OpenClaw, OpenCode). It intercepts tool calls—shell commands, URL fetches, file writes, package installs—and runs multiple threat checks (cloud URL reputation, 300+ local YAML heuristics, a prompt‑injection ML model, supply‑chain analysis, and Windows AMSI). If a request looks unsafe, Sage blocks it and logs the decision. Installation is done via the respective assistant’s marketplace or npm, and the project is open‑source under Apache 2.0.