conductor-oss/conductor

Conductor is an event driven agentic workflow engine providing durable and highly resilient execution engine for applications and AI Agents

What it solves

Conductor is a durable execution platform designed to manage complex, long-running workflows and AI agents. It prevents the loss of state during crashes, network failures, or restarts by persisting every step of an execution. It decouples the orchestration logic (the "graph") from the actual business logic (the workers), allowing developers to change the workflow's path without redeploying the worker code.

How it works

Workflows are defined as declarative, machine-readable JSON graphs. These graphs govern the sequence of tasks, including loops, branching, fan-out, and human approvals. Workers are polyglot—written in any language (Java, Python, Go, etc.)—and poll the server for tasks to execute and report results back. For AI agents, Conductor provides native LLM tasks and MCP (Model Context Protocol) tool discovery and calling, enabling agents to reason and act in durable loops.

Who it’s for

It is built for developers and operators who need to build scalable, reliable microservices orchestration or autonomous AI agents that require auditability, versioning, and the ability to recover from failures.

Highlights

  • Durable Execution: Every step is persisted, ensuring workflows survive infrastructure failures with configurable retries and timeouts.
  • AI Agent Orchestration: Native support for LLM chat completion, MCP tool calling, and vector workflows for RAG.
  • Polyglot Workers: Support for workers written in Java, Python, Go, JavaScript, C#, Ruby, or Rust.
  • Explicit Orchestration: Workflows are versioned JSON graphs, making the execution path inspectable and modifiable at runtime without code changes.
  • Execution Recovery: Ability to restart, rerun, retry, pause, resume, or terminate specific executions based on policy.

Related

  • Project
    vercel/workflow

    A TypeScript and JavaScript SDK that makes functions durable by persisting progress, retrying failed steps, and allowing workflows to suspend without using compute.

  • Project
    agentspan-ai/agentspan

    A durable runtime for AI agents that prevents execution loss during process crashes and supports long-running, event-driven workflows with human-in-the-loop approvals.

  • Project
    microsoft/conductor

    Conductor is a Microsoft‑maintained CLI that lets you define repeatable, deterministic multi‑agent AI workflows in YAML. It supports Copilot, Claude, and other providers, offers parallel execution, sub‑workflow composition, human‑in‑the‑loop gates, a real‑time web dashboard, and a terminal‑based fleet manager. Install via a one‑line script, run workflows locally or in CI, and keep everything version‑controlled.

  • Project
    langchain-ai/langgraphjs

    A low-level orchestration framework for building controllable, stateful AI agents with support for durable execution, long-term memory, and human-in-the-loop oversight.

  • Project
    chAng-L19/codex-redteam-mode

    A durable red-team runtime for Codex App and CLI that automates security workflows using goal contracts and evidence-based verification to prevent session interruptions and false completions.