Compozy Symbol
ALPHA
DocsInstallGetting StartedFeaturesPricing
CoreCLI ReferenceSchema DefinitionAPI Reference

Welcome

Welcome to the Compozy documentation! This guide will help you get started with building and managing reliable infrastructure for multi-agent AI systems, enabling scalable distributed workflows and production-ready automation.

Why Compozy?

Developing infrastructure for multi-agent AI systems presents unique distributed systems challenges that hinder scalability and reliability. As agents evolve from simple loops to complex workflows with fan-outs and sub-agent coordination, today's tools fall short in addressing production realities:

Distributed Complexity

Multi-agent systems require handling fan-outs across thousands of sub-agents, but developers spend 70% of their time managing throughput, fault tolerance, and cost control—turning scalable automations into fragile engineering nightmares.

Proprietary Limitations

Closed platforms restrict customization for agent prompts, untrusted context handling, and debugging, creating vendor lock-in that blocks innovation and escalates costs in high-volume scenarios like agentic map-reduce.

No Unified Standards

The absence of standardized protocols for agent orchestration leads to reinventing wheels—teams rebuild capabilities for monitoring, error recovery, and integration, as components from one system rarely compose with others.

Orchestration Overhead

Coordinating webhooks, APIs, rate limits, and sub-agent interactions demands extensive boilerplate, complicating deployment and maintenance of reliable, cost-efficient workflows at scale.

Solving these problems

Compozy provides open-source infrastructure that makes building and operating multi-agent systems routine and reliable, solving distributed challenges with high throughput, fault tolerance, and tools for effective management—empowering developers to focus on innovation:

Declarative Workflow Engine

Orchestrate multi-agent fan-outs using YAML for precise, LLM-friendly definitions of agents, tasks, and conditional logic, ensuring scalable and debuggable distributed execution.
See more

Multi-Provider Agent Support

Integrate 7 LLM providers (OpenAI, Anthropic, Google, etc.) with customizable prompts and memory, enabling seamless switching while handling untrusted context and monitoring.
See more

Advanced Orchestration

Support 8 task types for parallel fan-outs, routing, and collection processing, backed by Temporal for durable, cost-aware execution in high-volume multi-agent scenarios.
See more

Language-Agnostic Runtime

Build tools in TypeScript (via secure Bun runtime) with plans for more languages, providing isolated execution and permissions for reliable integration in agent workflows.
See more

MCP Protocol Integration

Connect to external services via Model Context Protocol with proxy support, enabling full manageability of workflows and sub-agents across AI ecosystems.
See more

Production Infrastructure

Enterprise-ready with PostgreSQL persistence, Redis memory, Temporal orchestration, authentication, and monitoring for scalable, debuggable multi-agent deployments.
See more

Getting Started

Ready to build infrastructure for multi-agent AI? Dive into our Quick Start Guide for your first workflow, or review Core Concepts to grasp Compozy's foundations for distributed agent systems.

Installation & Setup

Install Compozy CLI and set up your development environment for building AI-powered workflows