Streaming, stoppable
Tokens render as they arrive. Stop or regenerate any response mid-flight without losing the thread.
Nous Research · Hermes Agent
The control surface for your self-hosted Hermes agent.
A production-grade web client built for agentic work — threaded sessions, live streaming, transparent tool calls and reasoning traces. Runs today on a hosted model, and points at your Azure AI Foundry deployment the moment it is live.
- Sessions
- Threaded + resumable
- Transport
- Server-side streaming
- Target
- Azure AI Foundry
Tool calls in the open
Every tool invocation shows its name, arguments and result in a collapsible inline card.
Reasoning traces
Hermes' chain of thought is captured separately from the answer and stays folded until you want it.
Threaded sessions
Every conversation gets its own URL, auto-titled and restored on reload.
Swappable backend
One server adapter. Set your Azure AI Foundry endpoint and key, and traffic moves over — no UI rewrite.
Keys stay server-side
The browser never sees an inference credential; all model traffic goes through the app's own route.
Ready for the mobile app next
All model traffic flows through a single server route with a stable request contract, so the same backend serves this web console and a future native client without changes.
Start a session