Hermes Bots

I relied on Grok Bot, the hosted AI assistant app, and found it genuinely useful, but I ran through the free usage limit quickly. I pointed my self-hosted Hermes setup at doing the same job and built my own version: Hermes Bots, where I run Hermes (an LLM gateway and agent runtime) on a Hetzner server and reach it privately over Tailscale, with multiple bot profiles that each have their own personality, tools, and channels.

The interface centers on a Bots sidebar with search, per-bot threads, and group chats where several bots collaborate under a lead bot. Each bot gets its own integrations (Google Calendar, GitHub, Notion, Apify, Supabase via MCP, plus channels such as WhatsApp) and an avatar builder for shape, color, and hat styling. Sessions route models through multi-provider LLM routing (for example TokenRouter), so different bots can use the right model for the job.

Built as a solo full-stack and infrastructure project, the stack combines Hermes with FastAPI services, Docker, and Caddy on Hetzner, private access over Tailscale, and MCP-backed integrations. It shows how you can own the full assistant experience (profiles, tools, group orchestration, and deployment) without giving up control of data or infrastructure.