Deploy and Host Hermes Agent

Hermes Agent is an open-source AI assistant you run on your own infrastructure. hermes-station packages it into a single Railway-deployable container with a browser-based setup wizard — no config files to edit, no SSH required.

About Hosting

hermes-station runs as a single container backed by one Railway volume at /data. Everything — agent state, conversation history, credentials, memory — lives on that volume. Swap the container image for a new version and your data persists untouched.

The container exposes three surfaces:

/ — Hermes web chat UI
/admin — control plane: provider setup, channel management, gateway controls, live logs
/health — structured JSON healthcheck used by Railway's readiness probe

On first boot with no config, the container starts in degraded mode and walks you through setup via /admin. Nothing crashes — it just waits for you to add credentials.

Why Deploy

Privacy — your conversations and API keys never leave your infrastructure
Flexibility — swap LLM providers in seconds from the admin UI; no redeployment needed
Multi-channel — one agent identity across web, Telegram, Discord, Slack, email, and WhatsApp
Memory — holographic memory enabled by default; the agent remembers context across sessions
MCP tools — filesystem, GitHub, and web fetch tools pre-cached in the image, toggleable from /admin
Cost visibility — token spend shown in the UI by default so runaway loops are immediately visible

Common Use Cases

Personal AI assistant reachable from Telegram or Discord on your phone
Team bot connected to a shared Slack workspace
Self-hosted alternative to ChatGPT with your own Anthropic or OpenRouter key
Local-model gateway using Ollama or another OpenAI-compatible endpoint
Automated agent with scheduled tasks and long-term memory

Dependencies for hermes-station

hermes-station has no external service dependencies beyond the LLM provider you choose. Everything runs inside the container.

Dependency	Required	Notes
LLM provider API key	Yes	OpenRouter, Anthropic, OpenAI, Copilot, or custom
Railway volume	Yes	Mounted at `/data` — persists all agent state
`HERMES_ADMIN_PASSWORD`	Recommended	Locks `/admin`; open to anyone if unset
`HERMES_WEBUI_PASSWORD`	Recommended	Locks the web chat UI; open to anyone if unset
Telegram / Discord / Slack credentials	No	Only needed if you want those channels

Deployment Dependencies

Service	Purpose	Required
Railway Volume	Persistent storage for agent state, memory, and credentials	Yes
LLM Provider	Powers the agent — OpenRouter recommended for easiest setup	Yes
Telegram Bot	Reach the agent from Telegram	No
Discord Bot	Reach the agent from Discord	No
Slack App	Reach the agent from Slack	No

Setup after deploy

Set HERMES_ADMIN_PASSWORD and HERMES_WEBUI_PASSWORD in Railway variables
Open /admin on your deployed URL
Select a provider and paste your API key — OpenRouter gives access to all major models with one key
Optionally connect a channel (Telegram is the fastest to set up — just paste a bot token)
The agent starts automatically once a provider and at least one channel are configured