Agents

Type

Profile

Use key from

This is a local Hermes instance — your chats will use the running gateway's configured model and persona. Sessions are persisted on the Hermes side so multi-turn context survives page reloads.

Remote Hermes — you'll need the bearer key from the remote host's /root/.hermes/.api_server_key.

No agents yet. Click + Register above to add one.

Welcome to Agent OS

A dashboard for chatting with OpenAI-compatible agents and local Hermes instances. Register one on the left, then click into it to start a conversation.

Quick start

Local Hermes: select "Local Hermes agent" → pick a profile → Register. Alfred will be ready instantly.
Ollama / llama.cpp: start your server, then "OpenAI-compatible" with endpoint http://localhost:11434/v1.
Click the new agent in the sidebar to open chat

v0.2 — what's working:

OpenAI-compatible agents (Ollama, llama.cpp, vLLM, OpenAI)
Local Hermes agents with session continuity
Remote Hermes agents (manual endpoint + key)
Streaming chat (SSE over WebSocket) — works with all types
Image attachments (vision-capable models)
File uploads to /api/files/upload
Cookie-auth login + lockout

Coming in v0.3:

Builder tab (lovable/bolt-style)
Hermes skills/toolsets visible in chat UI

Shipped (Plan V1):

Voice: STT (faster-whisper) on uploaded audio + push-to-talk mic
Voice: TTS (edge-tts) auto-playback for agent replies (with 🔇 mute toggle)