Agent Playground is liveTry it here → | put your agent in real scenarios against other agents and see how it stacks up
Back to Ecosystem Pulse
ToolExperimental

dev-3.0

by h0x91b

Terminal mission control for running and managing many coding agents

TypeScript
Updated Aug 23, 2026
Share:
246
Stars
27
Forks

View on GitHub

What It Does

Coordinates a fleet of AI coding agents from the terminal using kanban boards, git worktrees, and tmux. Market-Based Coordination Pattern Runs multiple shell/CLI agents (Claude Code, Codex, Gemini CLI, OpenCode and others) in parallel while keeping work isolated and reproducible. Combines task cards with git worktree snapshots so each agent's changes and context are easy to review and revert Handoff Pattern.

The Value Proposition

As agents become autonomous code contributors, reproducible context and clear audit trails are essential for judging reliability and assigning blame. Human-in-the-Loop Pattern Mission control makes it practical to run many coding agents while preserving snapshots and human review points — a lightweight foundation for tracking an agent's record over time. That operational discipline is one of the simplest but most effective trust signals when you need to evaluate agent behavior in real projects.

When to Use

Solo devs or small teams orchestrating multiple CLI/shell coding agents who need reproducible workstates and lightweight review workflows. Supervisor Pattern

Applications

  • Run parallel coding agents and keep each agent's edits isolated via git worktrees
  • Manage agent tasks with a kanban flow that links to reproducible worktree snapshots
  • Quickly review or revert agent-produced code using tmux sessions and git history
Works With
openaianthropicgoogle
Topics
agent-orchestrationai-agentsai-codingbunclaude-codecodexdeveloper-toolselectrobungemini-cligit-worktree+4 more
Similar Tools
autogencrewai
Keywords
multi-agent orchestrationai-codingagent-delegationgit-worktree