Yuno: An AI Agent Orchestration Platform with Resumable Execution and Real-Time Monitoring

Yuno is a full-featured AI agent orchestration platform developed by pavankomateedi (GitHub repository: yuno-orchestration, released on 2026-06-03). It supports configurable agents, multi-agent workflows (including conditional branches and feedback loops), persistence and resumable execution based on LangGraph, Claude runtime, as well as real-time monitoring and communication functions. This project is not a proof of concept, but an engineering-level solution with complete front-end and back-end, deployment plans, and verification systems.

• Original Author/Maintainer: pavankomateedi
• Source Platform: GitHub
• Original Link: https://github.com/pavankomateedi/yuno-orchestration
• Release Time: 2026-06-03

Yuno aims to solve the problem of the lack of complete and reliable systems in the AI agent orchestration field, providing a production-grade solution with resumable execution and real-time monitoring capabilities.

Four-Layer Architecture

1. UI Layer: React + TypeScript + Vite (task control center, agent configuration, workflow builder, etc.)
2. Orchestration Backend: Python + FastAPI, dual-engine design (custom lightweight executor + LangGraph engine for persistence and recovery)
3. Runtime Layer: Claude API (replaces OpenClaw, as the latter's npm package is empty and its daemon is incompatible with stateless containers)
4. Persistence Layer: SQLite by default, Postgres/RDS optional (stores agents, workflows, execution history, and LangGraph checkpoints)

Security Design

Tool execution uses sandboxing: temporary directories, path restrictions, shell=False, environment cleanup; HTTP tools only allow GET requests and prevent SSRF.

1. Configurable Agents: Supports configuration of attributes such as personality, model, tools, memory, scheduling, skills, interaction rules, and guardrails
2. Multi-Agent Workflows: Conditional branches, feedback loops (e.g., coding → review → deployment), decision gating (pause to wait for human intervention)
3. Resumable Execution: LangGraph checkpoint mechanism allows recovery from the last state after crashes/timeouts
4. Real-Time Communication: WebSocket streaming for progress updates, Slack/WhatsApp chat with agents
5. Secure Tool Calls: Sandboxed execution with multiple protective measures to ensure safety

Verification System

• Backend: ruff code check, mypy type check, pytest tests, 6 golden scenarios verified by comparing dual engines
• Frontend: type check + production build

Deployment Plan

• Local Run: Single command to start backend (FastAPI) and frontend (Vite)
• AWS Deployment: CloudFront+S3 (SPA), ECS Fargate+ALB (FastAPI), EFS (SQLite persistence), Terraform infrastructure definition; optional RDS Postgres mode

Application Scenarios

• DevOps automation: complete workflow from coding → review → deployment
• Human intervention in decision-making: pause at key nodes to wait for human approval
• Multi-channel communication: Slack/WhatsApp interaction with agents

Project Value

Engineering completeness, architectural clarity, resumability, security, verifiability, deployability

Yuno is an engineering practice benchmark in the AI agent orchestration field, providing developers with clear architectural references, security practices, persistence solutions, verification systems, and deployment guidelines. Its dual-engine design, runtime abstraction, and verification system are particularly worth learning from, helping to build secure, reliable, and scalable multi-agent systems.