Agent Commander

Agent Commander is a mission-control dashboard for managing AI agent sessions across multiple hosts. It combines a Next.js dashboard, a Fastify control plane, and a Go-based agent daemon (agentd).

New in v0.2.0

• tmux-first manager — /tmux is now the primary day-to-day surface for live windows and panes across hosts
• Paperclip-inspired autonomy — autonomous orchestrators and workers with wake queues, governance approvals, budgets, concurrency policy, and runtime reuse
• Ruflo-inspired memory — repo and global memory for every session, plus trajectory capture and procedural knowledge
• Hermes integration — external wake triggers, summaries, and watchdog/digest hooks for the broader Heaviside automation stack

Screenshots

More screenshots

Sessions	Approvals

Session Detail	Visualizer

Quick Start

git clone https://github.com/cvsloane/agent-commander.git
cd agent-commander
pnpm install
pnpm dev

Copy and update environment files:

cp services/control-plane/.env.example services/control-plane/.env
cp apps/dashboard/.env.example apps/dashboard/.env

Run migrations:

pnpm db:migrate

Features

• tmux-first manager — work directly from /tmux with windows-first navigation and inline terminal control
• tmux-native sessions — discover and manage real tmux panes across hosts
• Interactive console — live streaming with full terminal control and read-only viewers
• Orchestrator + approvals — attention queue for input, approvals, and errors
• Autonomous agents — scheduled or manual orchestrators and workers with runtime reuse, preflight, and governance
• Scoped memory — repo and global memory for every session, with procedural knowledge and trajectory-backed learning
• Hermes-ready integration — service-auth wake triggers, webhook wakes, governance summaries, and external scheduler support
• Session generator — spawn multi-session templates with groups and links
• Visualizer — immersive full-screen activity view
• Search + analytics — global search and usage tracking
• Alerts — browser, audio, in-app toasts, and OpenClaw notifications
• Multi-host — one dashboard for many machines

Documentation

Guide	Description
Docs Index	Full documentation map
Overview	Concepts and workflows
Releases	Public release notes and milestones
Getting Started	Install and run locally
Tmux Manager	The tmux-first working surface
Sessions	Session lifecycle and actions
Console Streaming	Interactive tmux in the browser
Orchestrator	Attention queue and summaries
Automation and Memory	Autonomous runs, work queue, and scoped memory
Approvals	Permission requests and decisions
Coolify Deployment	Production setup on Coolify + Cloudflare
API Reference	REST + WS endpoints
Security	Security best practices

Public site:

• Feature site (apex) + docs site: docs/deployment-coolify.md

Architecture

API Reference (Summary)

• GET /v1/sessions, POST /v1/sessions/:id/commands
• GET /v1/approvals, POST /v1/approvals/:id/decide
• GET /v1/hosts, POST /v1/hosts
• GET /v1/repos, GET /v1/memory/search, POST /v1/memory
• GET /v1/automation-agents, POST /v1/automation-agents/:id/wake
• GET /v1/automation-agents/:id/preflight
• GET /v1/automation-runs, GET /v1/automation-runs/:id/events, GET /v1/automation-wakeups
• GET /v1/governance-approvals, POST /v1/governance-approvals/:id/decide
• GET /v1/work-items, POST /v1/work-items
• POST /v1/notifications/test

Security

• Set strong JWT_SECRET and NEXTAUTH_SECRET
• Use HTTPS in production
• Store secrets in a secret manager
• Rotate tokens on exposure

See SECURITY.md for reporting vulnerabilities.

License

MIT

Development Status

See project_status.md for recent development activity and context.