Hermes Agent deployment modes and Dashboard

Hermes supports deployment across all scenarios, from laptops to GPU clusters.

This chapter introduces the configuration methods, applicable scenarios, and comparisons of the six deployment modes one by one, as well as the visual management features of the Web Dashboard.

Detailed Explanation of Six Deployment Modes

Mode 1: Local CLI

The simplest deployment method: the Agent runs on the same machine as you, with zero configuration.

Direct dialogue:

hermes chat

TUI interface:

hermes --tui

Desktop application:

hermes desktop

Features: Simplest, zero configuration, suitable for development and personal daily use.

Mode 2: Gateway Service

Run the Agent as a 24/7 online Bot, always ready to respond to messages from messaging platforms, with scheduled task support.

Run in the foreground for debugging, logs output directly to the console:

hermes gateway

Register as a system service and start in the background for persistent production deployment:

hermes gateway install && hermes gateway start

Check gateway running status:

hermes gateway status

Other commands:

# 停止服务
hermes gateway stop
# 重启服务
hermes gateway restart
# 卸载系统服务
hermes gateway uninstall

Features: Agent stays online continuously, suitable for scenarios such as personal bots and team assistants.

Mode 3: Docker Container

The official image supports multi-Profile gateway, with processes managed by s6-overlay.

Pull official image:

docker pull nousresearch/hermes-agent:latest

Run (mount configuration directory):

docker run -d \
  --name hermes \
  -v ~/.hermes:/root/.hermes \
  -e OPENROUTER_API_KEY=sk-or-... \
  -e TELEGRAM_BOT_TOKEN=... \
  nousresearch/hermes-agent:latest

The multi-Profile gateway inside Docker is automatically supervised by s6-overlay.

hermes profile createIt automatically registers the s6 service slot, and after Docker restarts, it automatically restores all gateways that were running before.

Let the Agent's terminal commands also execute inside the container:

Features: Dual container isolation for processes and commands, suitable for CI/CD and production environments.

Mode 4: SSH Remote Server

The Agent runs locally, but all terminal commands are executed on the remote server.

Typical scenario: Send messages locally, and let the Agent train models and process large datasets on a remote GPU server.

Mode 5: Modal Serverless

The Agent executes commands in the Modal cloud sandbox (VM), incurring almost no cost when idle.

Features: Pay-as-you-go billing, suitable for occasional high-intensity computing tasks.

Mode 6: Daytona Managed Workspace

Both the Hermes process and the terminal sandbox are in the cloud, completely independent of the local machine.

Features: Fully cloud-based, managed cloud development environment.

Comparison of six modes

Web Dashboard

Dashboard is Hermes' visual management interface, which can manage the configuration, models, skills, MCP, and sessions of any Profile.

Start Dashboard

Start Dashboard using the following command:

hermes dashboard

The startup interface is as follows:

View skills:

Related commands:

# 启动 Dashboard（默认 Profile）
hermes dashboard

# 启动并预选指定 Profile
coder dashboard

# 指定端口
hermes dashboard --port 8080

Dashboard Feature Overview