Hermes Go

Hermes Go is the mobile client for a Hermes Agent that you run yourself: on your laptop, a VPS, or a homelab machine. Point it at your instance and your own agent is in your pocket, over a direct connection.

There is no relay server, no account to create, and no telemetry. The app opens a TLS connection to the address you give it and authenticates with your token.

WHAT YOU CAN DO

- Watch the work happen
Responses and reasoning stream in live over a direct WebSocket, with a separate drawer for the model's thinking. Tool calls appear as structured cards - Bash, Python, file edits, web lookups, MCP servers - so you can read what ran.

- Run agents as a team
Open several agents in their own rooms, each with its own soul, skills, and model, and hand tasks between them with @-mentions.

- Keep long tasks alive
Agent runs continue in the background while you use other apps, with a foreground notification showing what is running.

- Talk instead of type
Voice input, past sessions, and live server diagnostics sit in the same app as the chat.

- See the cost while it happens
Token counters and context-window meters show what a conversation is consuming, and you can switch models for quick work without leaving the thread.

- Make it yours
16 theme presets, adaptive dark mode, and a layout built for one-handed use.

BUILT FOR SELF-HOSTING

Hermes Go talks to your machine directly. You keep the models, the memory, and the logs - nothing routes through a middleman server.

REQUIREMENTS

- A running Hermes Agent instance your phone can reach (HTTPS, or a private network such as Tailscale)
- That instance's gateway URL and authentication token

Hermes Go is a client application. It needs the server: Hermes Go is the window, Hermes Agent is the engine.

ABOUT

Hermes Go is an independent, unofficial client for Hermes Agent, the open-source agent framework by Nous Research. It is not affiliated with or endorsed by Nous Research.