ORCHESTRATE v3.1 UAT — How AI Agents Tested Their Own Marketing Platform
We just shipped v3.1 of the ORCHESTRATE marketing platform and ran a full User Acceptance Test — not with human testers, but with the same AI agents that built it.
What is ORCHESTRATE?
ORCHESTRATE is a multi-channel content publishing platform that manages LinkedIn pages, Reddit posts, Dev.to blogs, YouTube uploads, and Printify merch — all from a single MCP (Model Context Protocol) server. It runs 100+ tools across 10 capability areas, orchestrated by AI agents following a rigorous agile methodology.
The UAT Process
For Sprint 13, we ran the platform through its paces:
- Blog Publishing — Created and published articles to Dev.to via MCP tools
- Trending Topic Discovery — Scanned Reddit for trending AI/automation topics
- Product Promotion — Pulled real product mockups from our Printify store and created targeted LinkedIn posts
- Video Production — Generated a narrated YouTube video entirely through AI: script generation, Piper TTS narration, ffmpeg video assembly, and YouTube upload
- Cross-Platform Distribution — Published the same content across LinkedIn, Reddit, Dev.to, and YouTube simultaneously
The Tech Stack
- Backend: Node.js REST API (port 3847) + MCP HTTP server (port 3848)
- Frontend: React + Vite + Tailwind
- AI Orchestration: MCP (Model Context Protocol) with 100+ registered tools
- Audio: Piper TTS sidecar (port 8500) for narration
- Video: ffmpeg in Docker for video assembly
- Merch: Printify API integration for product management
- Infrastructure: Docker Compose, single container deployment
Watch the UAT Video
We recorded the entire UAT experience as a narrated YouTube video:
The video was produced entirely by AI agents — from script generation to TTS narration to video assembly to YouTube upload.
What We Learned
Running UAT with AI agents revealed something interesting: the agents are better at systematic testing than ad-hoc exploration. They follow the acceptance criteria precisely, hit every endpoint, and document everything. But they don't stumble onto edge cases the way a human tester might.
The solution? Combine AI-driven systematic UAT with human exploratory testing. Let the agents handle the regression suite while humans focus on the unexpected.
Try It Yourself
The ORCHESTRATE platform is built on open protocols (MCP) and standard tooling. If you're building AI-powered content pipelines, the key insight is: treat your AI tools as first-class citizens in your CI/CD pipeline, not as ad-hoc helpers.
Shop merch: iamhitl.com
Follow the journey: I Am HITL on Dev.to
