About This Episode
I installed Hermes Agent on a Raspberry Pi 4 I've had sitting in a drawer for years, powered it off my router's USB port, and gave it a Telegram number. Now a $50 board generates images, videos and voice notes, and answers me from my phone. Full setup, start to finish: flashing the SD card, SSH, the one-line install, the setup wizard (including the one screen everybody gets wrong), API keys, picking a model that costs almost nothing, and the first message back. It idles at 600MB of 2GB RAM. The Pi does not care. What's in it: • Raspberry Pi Imager + which OS you actually need (64-bit, no exceptions) • Enabling SSH before you ever plug the thing in • Running the Pi off the router's USB port instead of a wall socket • The one-line Hermes Agent install • Local vs SSH in the wizard — the trick question • Telegram as your gateway so you're not stuck in a terminal • OpenRouter API key, credits, and reading the leaderboard • Owl Alpha and Tencent's Hy3 — the free models, and why they're free • Image generation via Nano Banana, video via fal.ai + LTX 2.3 • Free text-to-speech with Microsoft Edge TTS instead of paying ElevenLabs • Real cost per call, measured in the OpenRouter logs Tools used: Hermes Agent (Nous Research) — https://hermes-agent.nousresearch.com Raspberry Pi Imager — https://www.raspberrypi.com/software/ OpenRouter — https://openrouter.ai fal.ai — https://fal.ai Fair warning at the end of the video and here too: this is not a perfect tutorial. The Telegram bot creation step is not filmed. If you get stuck anywhere, paste the error into Claude, Codex, or Hermes itself — that is genuinely the workflow now. #raspberrypi #aiagent #hermesagent #nousresearch #openrouter #selfhosted #kimik3 #homelab #localai #telegrambot 0:00 A personal AGI on a $50 Pi 0:35 Flashing 64-bit Raspberry Pi OS 1:09 Wi-Fi vs LAN, and turning on SSH 1:35 Wiring it up — power from the router 2:07 First login and system update 2:53 One command installs Hermes Agent 3:17 Where the tokens come from 3:36 Local or SSH? (the trick question) 3:51 Telegram gateway + which tools to keep 4:35 Image generation + your OpenRouter API key 5:26 OpenRouter leaderboard: who runs what 6:11 Owl Alpha, the free mystery model 6:42 MiniMax M3 vs Tencent Hy3 7:15 What OpenClaw runs instead 7:34 "command not found" — and picking your LLM 8:17 Video generation with fal.ai + LTX 2.3 8:44 Hermes comes alive on the Pi 9:09 What it actually costs per call 9:35 htop: 600MB of 2GB used 10:00 The payoff — picture and voice note 10:22 Honest caveats