Agent Orchestration, Simplified

A declarative NoOps language for orchestrating distributed agent workflows with multi-language code generation and zero operational overhead

Get Started View on GitHub

example.kumeo

// Define a workflow for fraud detection
workflow FraudDetection {
  source: NATS("transactions")
  target: NATS("alerts")
  context: BayesianNetwork("risk.bn")

  agents: [
    LLM("ollama/llama3", prompt="Classify {{data}} as fraud? Context: {{context}}"),
    MLModel("isolation_forest.pkl", input=LLM.output),
    DecisionMatrix("policy.dmx", input=MLModel.output)
  ]
}

Key Features

Kumeo provides powerful tools to connect heterogeneous agents into coherent, event-driven workflows

Declarative Workflows

Define agent interactions as event-driven flows using a simple, readable syntax that focuses on what you want to achieve, not how.

Agent Types

Support for LLMs (Ollama/OpenAI), ML models (scikit-learn, ONNX), Bayesian networks, and human-in-the-loop integration.

Event Orchestration

Built on NATS for real-time, distributed communication between agents, with support for persistent streams and replay.

True NoOps

Write your workflow and run it - Kumeo handles everything from code generation to deployment, monitoring, and scaling with zero operational overhead.

Visual Editor

Svelte-based UI for workflow design and monitoring, making it easy to create and visualize complex agent interactions.

Multi-Language Generation

Automatically generates code in the most appropriate language for each component - Rust for performance-critical sections, Python for ML and Bayesian operations.

Architecture Overview

Kumeo is built on a modern, scalable architecture that connects heterogeneous agents through a central event system, with a true NoOps approach.

The compiler transforms declarative workflows into optimized multi-language code (Rust, Python, etc.) based on each agent's needs, and generates Kubernetes manifests that are automatically deployed to your infrastructure.

The platform manages the entire lifecycle - from code generation to deployment, monitoring, scaling, and updates - requiring zero operational overhead from developers.

Learn More