Hermes Console

What Hermes Console does

Hermes Console is a private, Android-first client for a Hermes Agent instance you control. It is a client, not an AI provider, so it needs a compatible Hermes Agent server and whatever model-provider access that server requires. Conversations support streaming Markdown and code, attachments, generated images, session history and model selection. Bots are kept separate from normal conversations, with profiles, mentions, rooms, tasks and activity.

An operations area covers runs and approvals, cron, Kanban, skills, memory, models and artifacts. A feature is shown only when the connected server exposes the capability it needs, so older servers degrade without invented endpoints. Voice offers dictation and a conversation mode with phone or Hermes-server speech routes and explicit fallbacks. Android extras include App Lock, read-only instances, notifications, widgets, a share sheet, QR pairing and connection diagnostics. The app has no XPeta Lab telemetry or advertising SDK, and credentials are protected with Android Keystore.

Key features

• Streaming chat with Markdown, code, attachments and session history
• Bots with profiles, mentions, rooms and tasks
• Runs, approvals, cron, Kanban, skills and memory controls
• Dictation and a voice conversation mode with explicit fallbacks
• QR pairing, App Lock, widgets and connection diagnostics
• No telemetry or advertising SDK, with credentials held in Android Keystore

When to use it

• Chatting with your own Hermes server from an Android phone
• Approving agent actions and checking cron or Kanban from the phone
• Pairing a phone to a server over a LAN or Tailscale with a QR code

Who it is for: Hermes Agent users with Android phones who run their own server and want a native mobile client.

How it fits with Hermes Agent

Built as a client for self-hosted Hermes Agent, following Hermes Desktop and Hermes Agent as its protocol contract.

Requirements: A compatible Hermes Agent instance on a server you control, reachable over HTTPS, a LAN or a private network such as Tailscale, plus any model-provider access that instance needs

FAQ