AgentApiary Documentation

Welcome to the AgentApiary documentation. Learn how to orchestrate AI agents with Kubernetes-inspired principles.

Architecture

Learn about AgentApiary's core components, data flow, and design principles.

API Reference

Complete REST and gRPC API documentation for managing AgentApiary resources.

CLI Reference

Command-line interface documentation for managing agents and resources.

Deployment

Guide to deploying AgentApiary in various environments and configurations.

Examples

Practical examples and use cases for getting started with AgentApiary.

What is AgentApiary?

AgentApiary is an AI agent orchestration platform inspired by Kubernetes. It brings declarative, cloud-native principles to AI agent coordination, enabling developers to deploy, scale, and manage complex multi-agent systems with ease.

Key Features

Declarative Configuration: Define your agent architecture with YAML/JSON manifests
Multi-Agent Orchestration: Coordinate specialized AI agents working together
Resource Management: CPU, memory, and task-tier based scheduling
Session Management: Track agent state and context across distributed systems
Message Bus: NATS-based pub/sub for agent-to-agent communication
Shared Memory: Session-scoped storage via Comb (shared memory)
Autoscaling: Automatically scale agents based on metrics
Guardrails: Token budgets, rate limiting, and safety controls

Core Concepts

AgentApiary introduces several core concepts:

Cells: Namespace isolation units (similar to Kubernetes namespaces)
AgentSpecs: Declarative specifications for how to run an agent
Hives: Collections of agents that work together
Drones: Running instances of agents
Sessions: Stateful contexts for agent interactions
Queen: The central orchestrator that reconciles desired state

Quick Start

Installation

Download the latest release from the releases page.

Linux / macOS

# Download the appropriate archive for your OS and architecture
tar -xzf apiary_1.0.0_linux_amd64.tar.gz

# Move binaries to your PATH
sudo mv apiaryd apiaryctl keeper /usr/local/bin/

Windows

# Extract the zip file
# Add the directory containing the .exe files to your PATH

Use Cases

Multi-Agent AI Systems: Coordinate specialized AI agents working together
AI Workflows: Build complex data processing pipelines with AI components
LLM Applications: Deploy and scale language model applications
Autonomous Systems: Orchestrate autonomous agents with shared context
Event Processing: Process events with AI-powered agents
Research Platforms: Scale AI research experiments across teams

Architecture

AgentApiary follows a modular, decoupled architecture adhering to SOLID and DRY principles. The platform consists of core components including the Queen (orchestrator), API Server, Scheduler, Launcher, Nectar (message bus), Comb (shared memory), Keeper (sidecar), and Store (BadgerDB).

Learn more in the Architecture Documentation.

Contributing

We welcome contributions! Please see our contributing guidelines on GitHub.

License

Licensed under the Apache License 2.0. See LICENSE for details.

Acknowledgments

AgentApiary is inspired by Kubernetes and brings the same declarative, cloud-native principles to AI agent orchestration.

Built with ❤️ for the AI community