conductor

Conductor is an open-source durable execution platform for microservices, AI agents and adaptive workflow graphs. Workflows are defined as graphs (JSON definitions) whose runtime choices, such as loops, branching, fan-out, tool calls, approvals, retries and cancellation, are executed durably by the Conductor server. It began at Netflix; the original Netflix/conductor repository is now archived, and conductor-oss is the actively developed open-source line.

It suits teams orchestrating many services and, increasingly, AI agents. Conductor has native LLM tasks, MCP tool calling, human approval steps and vector workflows for RAG, and its documentation shows an autonomous think-act agent loop that discovers tools via MCP, reasons with an LLM and repeats until done. It can be self-hosted with several persistence backends and message brokers, and a UI shows workflow executions.

Getting started: with Node.js 16+ and Java 21+, run `npm install -g @conductor-oss/conductor-cli` and `conductor server start`, then open http://localhost:8080. A Docker image is also available. Create your first workflow from the quickstart JSON with `conductor workflow create`.