Skip to content
Hermes SwitchUI

Your first chat

Your first chat

Open the app, pick a provider, and send your first message.

Before you start

You need both processes running:

  • Hermes Switch UI on http://localhost:3000
  • Hermes Agent on http://127.0.0.1:8642

For local development, pnpm dev starts the app and auto-starts the agent sidecar if it isn’t already running. If you used a different install method (Docker, Electron desktop, Unraid), follow that method’s launch instructions — see Install.

You also need an AI provider configured. If you haven’t done that yet, follow Connecting your AI provider first.

Quick steps

  1. Open http://localhost:3000 in your browser.

    If the app prompts for a password, enter the value of HERMES_PASSWORD set in your environment.

  2. Pick a theme (optional). The app ships with several themes; the default is fine for now.

  3. Make sure an AI provider is configured. If chat doesn’t respond when you send a message, see Connecting your AI provider.

  4. Click into the composer at the bottom of the chat screen and type a message.

  5. Press Send. You should see a streaming response and a new session appear in the sidebar.

[SCREENSHOT: composer first prompt, matrix-dark theme]

What you see

When you send a message:

  • The response streams in token-by-token in the chat view.
  • A new session is created automatically and shown in the sidebar on the left.
  • A context-window indicator updates as the conversation grows, so you can see how much of the model’s context budget you’ve used.
  • A model switcher near the composer lets you change which model handles the next turn.

[SCREENSHOT: streaming response, matrix-dark theme]

The session is saved locally. You can leave the page, come back later, click the session in the sidebar, and continue the conversation.

Details

  • Session history — All chats are listed in the left sidebar. Click any session to resume.
  • Model switcher — Switching the model mid-conversation works; the agent will use the new model for the next turn.
  • Context indicator — Shows tokens used out of the model’s context window. When it gets close to full, start a fresh session or compress the conversation.
  • Streaming — Responses stream over Server-Sent Events. If a stream stalls, the app surfaces a stalled state instead of hanging.

Common issues

  • Nothing happens when you press Send — Usually means no provider is configured or the agent is not reachable. See Connecting your AI provider and Agent won’t connect.
  • “Agent unavailable” banner — The app cannot reach the Hermes Agent on port 8642. See Agent won’t connect.
  • The response stops mid-stream — Usually a provider-side timeout or a transient network issue. Resend the message. If it keeps happening, check the agent logs in ~/.hermes/.