Conductor OSS
Conductor OSS is the Netflix-founded, Orkes-stewarded open source workflow and agentic AI orchestration platform. It provides a durable, event-driven workflow engine for coordinating microservices, long-running tasks, human approvals, and LLM-powered agents across any language or cloud, with first class support for HTTP, gRPC, Kafka, AMQP, SQS, and MCP-based tool calling.
APIs
Conductor OSS REST API
The Conductor OSS REST API is the canonical control plane for the workflow engine — registering and versioning workflow and task definitions, starting and querying workflow exec...
Conductor MCP Server
Conductor MCP is an official Model Context Protocol server for Orkes Conductor that exposes workflow execution, task management, and metadata operations as MCP tools so that LLM...
Collections
Features
Workflow state is persisted at every step so executions survive worker restarts, server failover, and long human-in-the-loop waits.
Tasks are executed by language-agnostic workers that poll the server, with official SDKs in Java, Python, Go, JavaScript / TypeScript, C#, Clojure, Ruby, and Rust.
Built-in HTTP, Kafka, JQ, wait, human approval, sub-workflow, fork, join, switch, do-while, and dynamic-fork operators compose workflows without writing code.
Native LLM tasks for 14+ providers, vector database integrations for RAG, MCP tool calling, and human-in-the-loop checkpoints make Conductor a runtime for agentic AI applications.
Event handlers consume from Kafka, AMQP, SQS, NATS, and webhook sources to start workflows or complete tasks from external systems.
Stateless API servers and worker pools scale horizontally, with pluggable persistence (PostgreSQL, MySQL, Cassandra, Redis) and indexing (OpenSearch / Elasticsearch).
Built-in metrics, structured logs, distributed tracing, and a Swagger UI on every server expose the workflow surface for operators.
Use Cases
Coordinate sagas, compensations, retries, and timeouts across distributed microservices without bespoke orchestration code.
Run production-grade agentic workflows with reasoning loops, tool calling, MCP integration, RAG retrieval, and human approvals.
Reliable long-running flows for payments, fulfillment, KYC, and onboarding — the reference example workload shipped by Conductor.
Schedule and orchestrate ETL, model training, and inference pipelines with durable retries and conditional branching.
Pause workflows on human tasks, capture approvals via UI or API, and resume execution deterministically.
Bridge message brokers and SaaS webhooks into deterministic workflows across the enterprise.
Integrations
System task and event-handler integration for Kafka topics.
Built-in AMQP system task and event source.
Native SQS event queue and system task module.
NATS event queue integration for triggering workflows.
Pluggable persistence backends for workflow and task state.
Pluggable indexing backends powering search and history.
External payload storage for large workflow inputs and outputs.
Official MCP server exposes Conductor to AI agents as tools.
gRPC API surface alongside the REST API for high-performance clients.
Solutions
Apache 2.0 open source workflow engine, self-hosted via Docker, Helm, or source build.
Free hosted Conductor sandbox by Orkes with all enterprise features enabled for prototyping and learning.
Fully managed Conductor on AWS, Azure, GCP, or on-prem with SOC2, RBAC, SSO, audit logs, and up to 99.99% availability SLA.