Show HN: Hermes WebTop ~ Experience Hermes Agent without Hassle Hermes-WebTop launches a browser-based Linux desktop with a pre-configured Hermes AI assistant that runs entirely through GitHub Codespaces, eliminating the need for local hardware or GPU. The tool provides a complete Ubuntu MATE environment, VSCode integration, and free LLM support via ModelRelay, allowing users to deploy a personal AI agent in under 10 minutes. This removes the primary barrier to using Hermes—a dedicated machine with a GPU—making the framework accessible for testing, student projects, and budget-conscious users. Run Hermes-Agent inside a browser-based VSCode + Linux desktop with free LLM support through ModelRelay. She will grow with you ... Hermes-WebTop gives you a fully functional Hermes AI assistant in your browser in under 10 minutes — no powerful PC, no Docker on your machine, no GPU required. Just open this repo in a GitHub Codespace and you get: - A complete Ubuntu MATE desktop WebTop for computer use - CodeServer at port 8888 with Hermes Extension installed and preconfigured - Ollama server pre-installed and auto-started - ModelRelay pre-installed, auto-started and pre-configured as default model - Hermes gateway accessible via desktop launcher Mnemon https://github.com/mnemon-dev/mnemon as your Hermes default memory provider https://github.com/gitricko/hermes-plugin-mnemon - Persistent volume for your config and settings When you're ready to go production, simply move the same Docker setup to your own machine or VPS. Hermes is an AI agent framework that connects LLMs directly to your communication platforms WhatsApp, Telegram, Slack, Discord, etc. and can run cron jobs, spawn sub-agents, speak/listen, and give you a beautiful dashboard. The only catch? You normally need a dedicated machine with GPU. Hermes-WebTop removes that catch completely. Perfect for: - Trying Hermes-Agent risk-free - Free LLM APIs through ModelRelay https://github.com/ellipticmarketing/modelrelay or OmniRoute https://github.com/diegosouzapw/OmniRoute - Students / hackers / evaluators - Anyone who wants a personal AI assistant without breaking the bank - Open this repository in a GitHub Codespace big green "Code" button → Codespaces → New It is recommended that you use 4 cpu core and 16G codespace. - In the Codespace terminal run: make start - Wait ~60 seconds after first launch. It takes about 5-7 minutes if docker image is not loaded yet. When the web desktop URL appears in the Codespace Ports tab, click it. Demo on pre-installed WebApps - Use VSCode on Web Code-Server to interact wit your agents: - Interact with files system using VSCode on the browser - Interact with Hermes Agent using VSCode Extension OR via hermes-cli in the VSCode Terminal - Claude Code + Hermes both using Terminal-Cli or VSCode Extension shared the persistent memory via mnemon https://github.com/mnemon-dev/mnemon so that Hermes remembers what you did in Claude and vice-versa. - Use WebTop interface in port 3000 when you need to do computer-use - Configure your Hermes with your favourite messaging app - Recommendation: Watch Youtube ;- - Last NOTE: - For localhost install.. just use http://localhost: 3000,8888,9119,7352,20128 http://localhost:%5B3000,8888,9119,7352,20128%5D - For localhost install.. just use Zero local install — everything runs in browser via GitHub Codespaces Free-tier friendly — uses ModelRelay, Ollama daily cloud credits or NVIDIA Build API fallback Persistent config — docker volume backup and restore after Codespace recreation Easy backup/restore — make backup / make restore One-command everything — powerful Makefile + clean docker-compose.yml Auto-start ModelRelay — Default configuration for Free LLM API to Hermes Auto-start OmniRoute — You need some configuration before it work but it is flexible and powerful Auto-start Ollama — custom init script on WebTop boot Colima / local Docker support ready Built-in code-server IDE — browser-based VS Code on port 8888 Multiple AI VSCode Extension preinstall/config with ModelRelay - Cline, Hermes and ClaudeCode Mnemon as default memory provider - a knowledge store for Hermes with intent-aware recall, importance decay, and auto-deduplication This image includes code-server and exposes it on port 8888 . code-server is installed automatically in the container.- The desktop launcher CodeServer starts it inside the WebTop environment. - In Codespaces, use the forwarded private port 8888 . - Locally, open http://localhost:8888 . Note: this setup may use code-server --auth none in development, so keep port 8888 private. For local production use, secure it with an authenticated reverse proxy or firewall. - Hermes Agent's Extension is preinstalled and configured in VSCode - Cline Extension is also preinstall and configured to ModelRelay - Start Hacking away in VSCode, use WebTop if you need to monitor agent do desktop-use operations. eg: Non-Headless Chrome debugging for instance / Linux Computer-Use The WebTop URI is automatically protected — no one else can reach it. GitHub Codespaces forwards ports privately by default this is the setting the make start command uses . According to official GitHub documentation https://docs.github.com/en/enterprise-cloud@latest/codespaces/reference/security-in-github-codespaces : "All forwarded ports are private by default, which means that you will need to authenticate before you can access the port." "Privately forwarded ports: Are accessible on the internet, but only the codespace creator can access them, after authenticating to GitHub." - The URL you click in the Ports tab https://