Skip to content

Hermes Stack — Self-hosted AI Brain

Hermes Stack is the project that ties my whole AI infrastructure together. It packages Hermes Agent, 9Router and OmniRouter into a single Hugging Face Space that deploys in minutes and costs nothing. One click, and you own a cloud-hosted AI brain: an agent runtime, a model router and a console to drive both.

What is inside

ComponentRole
Hermes AgentThe brain — plans tasks, calls tools, runs automations, remembers context.
9RouterMulti-provider model routing for nine upstream families.
OmniRouterMy Go router adding weights, quotas and failover on top.
HF SpaceFree hosting with zero cold-start pain for this class of workload.

Why it matters

Most "AI assistant" setups are either a hosted SaaS you don't control or a local rig you can't reach from your phone. Hermes Stack is the third option: self-hosted, cloud-reachable, and free. I use it as the coordinator for my other agents — it holds API routes, serves as a webhook brain for automations, and gives every tool in my chain a stable, roammable endpoint.

  1. Fork the Space (or deploy from the repo with one command).
  2. Paste your upstream keys — or point it at free bridges like GLM Free API.
  3. Talk to it from Telegram, curl, or any OpenAI-compatible client.

The Agentic Coding section covers how Hermes drives the rest of my toolchain day to day. This portfolio itself was coordinated through it — briefs, reviews and deploys all routed through a stack that costs less than a coffee. Which is to say: exactly nothing.