AgentLoop Documentation
Welcome to the AgentLoop documentation. AgentLoop is an autonomous agent orchestrator that coordinates multiple AI agents working together to complete complex software development tasks.
Quick Navigation
Install AgentLoop and run your first command in minutes.
Getting StartedLearn how to use the interactive TUI for conversational development.
Interactive ModeUnderstand the difference between natural language and system commands.
Commands vs Slash CommandsExplore how agents, DAGs, and the orchestrator work together.
ArchitectureWhat is AgentLoop?
Agentloop is a neuro-symbolic, deterministic AI agent engine that generates verifiably correct, production-grade code and autonomously executes complex real-world workflows.
Built for machine learning and real-time predictive analytics, it powers software development, ML pipelines, DevOps pipelines, AI assistants, and autonomous control systems (including drone flight control), delivering mission-critical enterprise systems with unmatched speed and reliability. Specialized agents work together seamlessly:
- Orchestrator Agent - Manages overall workflow and delegates tasks
- Product Manager Agent - Breaks down projects into AGILE tasks and user stories
- Engineer Agent - Autonomous code implementation with proper git workflows
- QA Tester Agent - Automated testing, issue identification, and validationn
- Code Analyzer Agent - Deep codebase analysis for architecture understanding
Key Features
| Feature | Description |
|---|---|
| Multi-Agent Orchestration | Coordinate multiple specialized AI agents working together |
| DAG-Based Task Management | Define dependencies and let the orchestrator optimize execution order |
| Interactive Kanban Board | Visual task management in your terminal |
| Parallel Execution | Multiple agents work on independent tasks simultaneously |
| Git Worktrees | Isolated development environments for true parallelism |
| Container Sandboxing | Optional Podman-based isolation for secure agent execution |
| Self-Healing Recovery | Automatic detection and recovery from failures |
| Jira Integration | Bidirectional sync with your existing Jira workflows |
Documentation Sections
Getting Started
- Installation - Install AgentLoop globally or locally
- Interactive Mode - Working with the TUI
- Commands vs Slash Commands - Understanding the two interaction modes
Core Concepts
- Architecture - System design and agent flow
- Orchestrator - Task scheduling and DAG management
- Agents - Agent types and their capabilities
- Kanban Board - Visual task management
Advanced Features
- Git Worktrees - Parallel development with isolated branches
- Container Sandboxing - Podman-based agent isolation
- Background Daemon - Persistent background execution
Integrations & Configuration
- Integrations - Jira, GitHub, and MCP servers
- Configuration - TOML configuration reference
- Commands Reference - Complete command listing
Need help? Join our Discord community or visit agentloop.run .