tmux-ide

Turn any project into an autonomous multi-agent development environment. One config file, multiple AI agents, mission-driven orchestration.

┌─────────────┬─────────────┬─────────────┐
│             │             │             │
│  Lead       │  Frontend   │  Backend    │  ← skill-matched agents
│  (Planner)  │  (React)    │  (API)      │
│             │             │             │
├──────┬──────┴─────────────┴──────┬──────┤
│      │                           │      │
│ Dev  │      Mission Control      │ Shell │
│ Srv  │      (TUI dashboard)      │      │
└──────┴───────────────────────────┴──────┘

Two Modes

Classic Mode -- Define panes in ide.yml, run tmux-ide, get a terminal IDE with labeled panes and dev servers.

Missions Mode -- Autonomous multi-agent orchestration. Define a mission, and agents self-organize through milestones, validation contracts, and skill-based dispatch.

Mission → Milestones → Tasks → Agents (skill-matched)
                                 ↓
                           Validation → Remediation
                                 ↓
                           Mission Complete → PR

Quick Start

npm i -g tmux-ide

# Classic mode
tmux-ide init           # scaffold ide.yml
tmux-ide                # launch

# Missions mode
tmux-ide init --template missions
tmux-ide mission create "Build auth system" -d "JWT + refresh tokens"
tmux-ide                # → agents, dashboard at localhost:6060, autonomous

Feature Highlights

Milestone Gating

Sequential execution phases. Tasks only dispatch when their milestone is active and all predecessors are done. Auto-progression when a milestone's tasks complete.

Validation Contracts

Assertion-based verification. Define testable claims in a contract. A validator agent independently checks work. Failed assertions auto-create remediation tasks.

Skill-Based Dispatch

Match task specialty to agent capabilities. Specialist agents get specialist tasks. Generalists pick up unspecialized work. Tasks wait for the right agent instead of dispatching to the wrong one.

Knowledge Library

Shared learnings auto-append as agents complete tasks. Architecture docs, service registries, and tag-matched references are injected into dispatch prompts.

Live Dashboard

Web dashboard with KPIs, milestone timeline, agent utilization, validation status, and activity timeline. Auto-refreshes every 5 seconds.

Researcher Agent

Continuous internal auditing. Triggers on mission start, milestone progress, stalls, retry clusters, and discovered issues. Writes findings to the knowledge library.