Mon setup d’IA agentique : comment tout fonctionne ensemble

Mon setup d’IA agentique : comment tout fonctionne ensemble

My agentic AI setup: how everything works together

Title history
  • Mon setup d’IA agentique : comment tout fonctionne ensemble — observed on 20261111-0817 UTC (YouTube)

by Rudy IO · View the channel on PSI

Applied Sciences & Engineering AI — News & Applications UYComputer scienceUYQArtificial intelligence (AI)
🎙 Rudy IO 👥 4K 📅 September 16, 2026 ⏱ 26 min 👁 52K 📄 expert opinion 🧭 2026-11-11
Available in: English (current) Français

Keywords

agentic AIsetupMCPself-hostedautomation

Summary

In this video, Rudy IO presents his personal agentic AI setup, explaining how he uses AI agents to manage his daily tasks. The core of the setup is a powerful PC running three agents: ChatGPT Codex, Claude, and Hermes (an open-source harness). These agents share context through Markdown files (agent.md and project folders) and have access to various tools and software, including video generation via ComfyUI, DaVinci Resolve, and Blender. They can control the machine, execute commands, and manage remote services via SSH. The author also describes using a local server (a small Firebat N100) for self-hosted services like Home Assistant and media tools, and a remote VPS hosted on Hostinger for externally accessible services. Cloudflare is used for DNS routing and security, including DDoS protection and two-factor authentication. API keys are protected using Infisical, a secrets management service, to prevent agents from directly seeing sensitive credentials. The video includes a concrete example of building an application from a voice command, and discusses the trade-offs of hosting choices, maintenance, and the importance of keeping human validation for critical actions.

179 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video offers substantial practical value for those interested in building their own AI agent infrastructure. It provides a concrete, detailed architecture that combines local and cloud resources, with clear explanations of why each component is chosen (e.g., local server for control, VPS for reliability). The argumentation is solid and experience-based: the author explains the reasoning behind each decision, such as using Infisical for secrets to mitigate agent compromise, and keeping human approval for sending emails. The discussion of trade-offs, such as the choice to host client services on Hostinger rather than his own hardware, shows a balanced perspective. However, the argumentation relies solely on personal experience without external validation or references, which limits its generalizability but strengthens its authenticity as a real-world case study.

Scientific Rigor, Source Quality, Title Accuracy

The video is scientifically rigorous in its transparency about the architecture and potential weaknesses, but it does not cite external sources or studies. The quality of sources is limited to the author’s own setup and the official documentation of tools (implicitly referenced). The title accurately matches the content: the video is indeed a walkthrough of how the agentic AI setup works together. Since the video has chapters provided by the creator, the timestamped moments are left empty. No public comments were provided for analysis.

224 words

Title / Content Match

The title accurately reflects the content: the video explains how the various components of the agentic AI setup work together.

Quality & Reliability

7/10

The video provides a detailed and honest walkthrough of the author's personal AI agent setup, including security considerations and limitations. Claims are based on personal experience, not peer-reviewed research, hence a moderate-high reliability score.

Chapters

Cited Sources

  • NXUS — Official website and resources of the author, mentioned in the description.

Contribution & Novelties

This video contributes a rare, in-depth personal case study of a fully agentic AI infrastructure, showcasing how multiple AI agents can be orchestrated to share context and execute complex tasks across local and cloud services. It offers a pragmatic blueprint for enthusiasts and professionals, emphasizing security via secrets management and Cloudflare, and explaining the rationale behind self-hosting versus VPS. The original insight lies in the seamless integration of open-source tools (Hermes, ComfyUI) with commercial services (ChatGPT, Claude) and the practical handling of API key protection.

Pour aller plus loin :

  • Model Context Protocol (MCP) — Official protocol for connecting AI models to external tools and data, a foundational element mentioned in the video.
  • Self-hosting — Wikipedia article explaining the concept of self-hosted services, relevant to the local server discussion.
  • Cloudflare — Reference for the CDN, DNS, and security services used to protect the exposed containers.
  • Home Assistant — Open-source home automation platform that the author runs as a self-hosted service.

160 words

Radar Profile

The profile shows a balanced high performance across quantity and quality of information, with a strong technical depth. The fiabilite_globale score is relatively high for a non-academic source, reflecting the honest and transparent approach of the author.

Reliability 7/10