AI Fleet

ai-fleet 🚀

*"Spin up and command a fleet of AI developer agents from your terminal."*

Table of Contents

- Quick Start
- Installation
- Core Features
- Commands Reference
- Configuration
- Development Setup
- Examples
- Troubleshooting
- Product Requirements

Quick Start

Installation

From PyPI (when published)

From Source (Development)

See Development Setup section below.

Key Concepts

Project-Based Configuration

AI Fleet works on a per-project basis. Each git repository you want to use AI Fleet with needs to be initialized:

1. One config per project: Configuration lives in .aifleet/config.toml within your project
2. Automatic detection: Commands work from any subdirectory within your project
3. Isolated state: Each project maintains its own agent state and worktrees
4. Team friendly: Configuration can be committed and shared with your team

Core Features

🎯 Single Agent Management
- Create individual agents with custom prompts
- Each agent runs in an isolated git worktree
- Full tmux session management
- Real-time CPU/memory monitoring

🚀 Parallel Agent Orchestration
- Fanout: Launch N agents with the same prompt
- Multi: Launch multiple agents with different tasks
- Batch management and cleanup
- Pattern-based operations (glob support)

🔧 Project Integration
- Automatic credential file copying
- Custom setup commands per project
- Git worktree isolation
- Branch naming conventions

📊 Monitoring & Control
- Real-time agent status with CPU/RAM usage
- Log tailing and session attachment
- Interactive prompt sending
- Batch operations

Commands Reference

Core Commands

fleet create  [--prompt PROMPT]
Create a new agent on a specific branch.

Options:
- --prompt, -p: Initial prompt to send to the agent
- --agent, -a: AI agent to use (default: claude)
-…