# Orloj

An open-source orchestration runtime that treats AI agents, tools, and policies as declarative YAML resources, scheduling, routing, and governing them like production infrastructure.

Orloj is an orchestration runtime for multi-agent AI systems that declares agents, tools, models, memory, schedules, webhooks, approvals, and policies as versioned YAML resources with desired state, status, and controllers, in the way infrastructure platforms treat services. It provides sequential and message-driven execution, workers, leases, heartbeats, retries, idempotency keys, and dead-letter states for handling failures in agent workflows.

Interfaces include YAML manifests, a command-line tool called orlojctl, a REST API, SDKs, Kubernetes custom resource definitions, and a web console. Its model and context layer handles provider routing, fallback models, secrets, and token budgets, while its tool layer adds isolation, authorization, approval gates, retry policy, and audit-friendly traces around tool calls that reach production systems.

It is written in Go, released under the Apache-2.0 license, and is under active development, so its APIs and resource schemas may change before a 1.0 release.

Website: https://orloj.dev