AI Agent Orchestration Platform
Full-stack agent orchestration platform with a visual workflow builder, real-time monitoring, and Telegram integration — letting users create AI agents, configure their behavior, and connect them into collaborative multi-step pipelines.
AI Agent Orchestration Platform
Role: Full-Stack AI Engineer
Program: Gauntlet AI — 2-Month Immersive for AI Engineers
Live Demo: agentorchestration.vercel.app
GitHub: github.com/alediez2048/yuno_agent_platform
Tools: Next.js 14, TypeScript, React Flow, SQLite, WebSocket, Anthropic Claude, Telegram (telegraf), Tailwind CSS, shadcn/ui
Overview
A full-stack AI agent orchestration platform inspired by payment orchestration patterns — smart routing, unified gateways, real-time monitoring, and autonomous operations — applied to AI agents instead of payment processors.
Users can create AI agents with custom roles, system prompts, and guardrails, then connect them into collaborative workflows using a visual drag-and-drop DAG editor. The platform includes real-time execution monitoring, Telegram integration, and pre-built workflow templates.
Key Features
- Agent Management: Create and configure AI agents with custom roles (Researcher, Analyst, Coder, Reviewer, Deployer), system prompts, model selection, tool access, and guardrails
- Visual Workflow Builder: Drag-and-drop DAG editor (React Flow) to connect agents into multi-step collaborative pipelines with feedback loops
- Real-Time Monitoring: WebSocket-powered dashboard tracking agent execution, token usage, and costs
- Telegram Integration: Interact with agents via Telegram bot with slash commands and message routing
- Pre-Built Templates: Competitive Intelligence, Market Expansion, Dev Pipeline, and Research Pipeline workflows
- Export & Share: Copy markdown, download
.md, print/PDF, and shareable links for execution results - In-App Assistant: Command center using platform-aware context for onboarding and help
Architecture
- Next.js 14 App Router with co-located API routes and frontend
- React Flow for the visual graph/workflow builder
- SQLite (WAL mode) via better-sqlite3 for zero-infrastructure persistence
- WebSocket for bidirectional real-time communication
- Anthropic Claude as the primary LLM provider, configurable per agent
- Zod for runtime schema validation across all API inputs
Tech Stack
| Layer | Stack |
|---|---|
| Language | TypeScript 5.x (shared types across frontend and backend) |
| Framework | Next.js 14 (App Router) |
| Workflow Builder | React Flow 11.x |
| Database | SQLite (WAL mode) via better-sqlite3 |
| Real-Time | WebSocket via ws |
| AI Runtime | Anthropic Claude API |
| Messaging | Telegram via telegraf |
| UI | Tailwind CSS 3.x, shadcn/ui (Radix-based) |
| Testing | Vitest (unit, integration, smoke tiers) |
| Validation | Zod 3.x |
