multi-agent-ff15

OpenCode Multi-Agent Command System

5 AI agents running in parallel with a single command

Noctis (King) commanding 3 Comrades (Ignis, Gladiolus, Prompto) with Lunafreya (Oracle) operating independently - actual session screenshot

With a single command, Noctis (King) directly assigns tasks to 3 Comrades (Ignis, Gladiolus, Prompto) who execute in parallel. Meanwhile, Lunafreya (Oracle) operates independently, consulting directly with you and commanding Noctis when needed. All agents run as independent OpenCode processes in tmux. Communication flows through YAML files and tmux send-keys, meaning zero API calls for agent coordination.

Framework: Built on OpenCode

Why Noctis?

Most multi-agent frameworks burn API tokens on coordination. Noctis doesn't.

	OpenCode	LangGraph	CrewAI	multi-agent-ff15
Architecture	Agents with tools	Graph-based state machine	Role-based agents	Feudal hierarchy via tmux
Parallelism	Limited	Parallel nodes (v0.2+)	Limited	5 independent agents
Coordination cost	API calls	API + infra (Postgres/Redis)	API + CrewAI platform	Zero (YAML + tmux)
Observability	Logs only	LangSmith integration	OpenTelemetry	Live tmux panes + dashboard
Skill discovery	None	None	None	Bottom-up auto-proposal
Setup	CLI install	Heavy (infra required)	pip install	Shell scripts

What makes this different

Zero coordination overhead — Agents communicate through YAML files on disk. The only API calls are for actual work, not orchestration. Run 5 agents and pay only for 5 agents' work.

Full transparency — Every agent runs in a visible tmux pane. Every instruction, report, and decision is a plain YAML file you can read, diff, and version-control. No black boxes.

Battle-tested hierarchy — The Noctis → Comrades chain of command prevents conflicts by design: clear ownership, dedicated files per agent, event-driven communication, no polling. Lunafreya operates independently outside this hierarchy.

Bottom-up skill discovery — As Comrades execute tasks, they automatically identify reusable patterns and propose them as skill candidates. You decide what gets promoted to a permanent skill.

🚀 Quick Start

🪟 Windows Users (Most Common)

Step 1	📥 Download the repository Download ZIP and extract to `C:\tools\multi-agent-ff15` Or use git: `git clone https://github.com/atman-33/multi-agent-ff15.git C:\tools\multi-agent-ff15`
Step 2	🖱️ Run `install.bat` Right-click → "Run as administrator" (if WSL2 is not installed). Sets up WSL2 + Ubuntu.
Step 3	🐧 Open Ubuntu and run the following (first time only) cd /mnt/c/tools/multi-agent-ff15 ./first_setup.sh
Step 4	✅ Stand by Me! ./standby.sh

🔑 First time only: Authentication

After first_setup.sh completes, run the following once to authenticate:

# 1. Apply PATH changes
source ~/.bashrc

# 2. Start OpenCode
opencode
#    → Select your preferred AI model provider
#    → Follow authentication prompts
#    → Type /exit to quit

Authentication is saved to ~/.opencode/ and won't be needed again.

📅 Daily startup (after initial setup)

Open Ubuntu terminal (WSL) and run:

cd /mnt/c/tools/multi-agent-ff15
./standby.sh

🐧 Linux / Mac Users (click to expand)

Initial Setup

# 1. Clone the repository
git clone https://github.com/atman-33/multi-agent-ff15.git ~/multi-agent-ff15
cd ~/multi-agent-ff15

# 2. Grant execution permission to scripts
chmod +x *.sh

# 3. Run initial setup
./first_setup.sh

Daily startup

cd ~/multi-agent-ff15
./standby.sh

❓ What is WSL2? Why is it needed? (click to expand)

About WSL2

WSL2 (Windows Subsystem for Linux) is a feature that allows you to run Linux inside Windows. This system uses tmux (a Linux tool) to manage multiple AI agents, so WSL2 is required on Windows.

If you don't have WSL2 yet

No problem! Running install.bat will:

Check if WSL2 is installed (auto-install if not)
Check if Ubuntu is installed (auto-install if not)
Guide you to the next step (how to run first_setup.sh)

Quick install command (run PowerShell as administrator):

wsl --install

Then restart your computer and run install.bat again.

✅ State after setup

6 AI agents will auto-start:

Agent	Role	Count
👑 Noctis	King - receives your commands and manages tasks	1
🌙 Lunafreya	Oracle - operates independently & commands Noctis	1
⚔️ Comrades (Ignis, Gladiolus, Prompto)	Workers - execute tasks in parallel	3
🌸 Iris	Guardian - monitors dashboard & notifies Noctis	1

tmux session: ff15 - unified session (6 panes)

📖 Basic Usage

Step 1: Connect to ff15 session

After running standby.sh, all agents automatically load their instructions and are ready to work.

Open a new terminal and connect to the ff15 session:

ffa    # Alias (tmux attach-session -t ff15)

Step 2: Give your first command

Noctis is already initialized! Just give a command:

Research the top 5 JavaScript frameworks and create a comparison table

Noctis will:

Write tasks to YAML files
Notify Ignis (manager)
Return control to you immediately (no waiting!)

Meanwhile, Ignis distributes tasks to Comrades and executes in parallel.

Step 3: Check progress

Open dashboard.md in your editor to see real-time status:

## In Progress
| Worker | Task | Status |
|--------|------|--------|
| Gladiolus | React research | Running |
| Prompto | Vue research | Running |
| Lunafreya | Angular research | Complete |

✨ Key Features

⚡ Parallel Execution

Generate up to 3 parallel tasks with a single command — results in minutes, not hours.

🔄 Non-Blocking Workflow

Noctis delegates instantly and returns control to you. No need to wait for long tasks to complete.

🧠 Cross-Session Memory (Memory MCP)

AI remembers your preferences across sessions. Tell it once, it remembers forever.

📡 Event-Driven (No Polling)

Agents communicate via YAML files and wake each other with tmux send-keys. No wasted API calls in polling loops.

📸 Screenshot Integration

Set screenshot folder in config/settings.yaml and tell Noctis "Check the latest screenshot" — AI immediately reads and analyzes it.

🛠️ Bottom-Up Skill Discovery

Comrades automatically identify reusable patterns and propose them as skill candidates. You approve what gets promoted to permanent skills.

🌍 Practical Examples

Example 1: Research Task

You: "Research the top 3 AI coding assistants and compare"

Executed process:
1. Noctis assigns to each Comrade:
   - Ignis: Research GitHub Copilot
   - Gladiolus: Research Cursor
   - Prompto: Research OpenCode
2. All 3 research simultaneously
3. Results aggregated in dashboard.md

Example 2: PoC Preparation

You: "Prepare PoC for project on this Notion page: [URL]"

Executed process:
1. Noctis fetches Notion content via MCP and assigns to each Comrade
2. Ignis: List items to verify
3. Gladiolus: Research technical feasibility
4. Prompto: Create PoC plan document
5. All results aggregated in dashboard.md, ready for meeting

📚 Documentation

For detailed information, see the docs folder:

Architecture - System design and communication protocol
Philosophy - Core principles and design decisions
Advanced Usage - Script options, workflows, and customization
Troubleshooting - Common issues and solutions
Mobile Access - Command from your phone
MCP Setup - Model Context Protocol configuration
Project Management - Managing multiple projects

⚙️ Configuration

Language Settings

Edit config/settings.yaml:

language: ja   # Japanese only
language: en   # Japanese + English translation

Party Formations

./standby.sh                # Normal formation (default)
./standby.sh --fullpower    # Full Power formation (premium models)
./standby.sh --lite         # Lite formation (budget mode)

📚 tmux Quick Reference

Command	Description
`ffa` (alias)	Connect to ff15 session
`Ctrl+B` then `0-5`	Switch between panes
`Ctrl+B` then `d`	Detach (keeps running)
`tmux kill-session -t ff15`	Stop ff15 session

Mouse operations enabled by default: scroll, click to switch panes, drag borders to resize.

Contributing

Issues and pull requests are welcome.

Bug reports: Create issue with reproduction steps
Feature ideas: Propose in Discussion first
Skills: Skills are personal by design and not included in this repo

🙏 Credits

This project is based on multi-agent-shogun by @yohey-w. We deeply appreciate the original work and the foundation it provided for this FF15-inspired multi-agent system.

📄 License

MIT License - See LICENSE for details.

Command your AI army. Build faster.

Name		Name	Last commit message	Last commit date
Latest commit History 240 Commits
.github/workflows		.github/workflows
.opencode		.opencode
assets		assets
config		config
desktop		desktop
docs		docs
scripts		scripts
templates		templates
tests		tests
.gitattributes		.gitattributes
.gitignore		.gitignore
AGENTS.md		AGENTS.md
CHANGELOG.md		CHANGELOG.md
LICENSE		LICENSE
README.md		README.md
README_ja.md		README_ja.md
first_setup.sh		first_setup.sh
install.bat		install.bat
package-lock.json		package-lock.json
package.json		package.json
standby.sh		standby.sh

License

atman-33/multi-agent-ff15

Folders and files

Latest commit

History

Repository files navigation