Messaging Platforms: Run Hermes on Telegram, Discord, Slack and 15+ More

Hermes on Your Messaging App#

The terminal is great for developers. Everyone else wants Hermes inside the app they already use. Hermes ships a messaging gateway that simultaneously drives 15+ platforms from a single process — Telegram, Discord, Slack, WhatsApp, Signal, Feishu, DingTalk, WeCom, QQ Bot, iMessage via BlueBubbles, and more. This lesson covers the gateway architecture, the shared security model, and the platform-by-platform capability map so you can pick the right channel.

Primary source: the messaging index at https://hermes-agent.nousresearch.com/docs/user-guide/messaging/. Each platform has its own page — this lesson is the cross-cutting orientation.

One Gateway, Many Platforms#

Rather than running one bot process per platform, Hermes uses a single gateway service that manages every configured channel at once — message routing, session persistence, cron execution, and voice delivery all flow through the same infrastructure. The practical consequence: adding a second or third platform doesn't double your operational burden.

The relevant commands live under hermes gateway:

hermes gateway setup     # interactive setup for a new platform
hermes gateway run       # run in foreground (useful for debugging)
hermes gateway start     # start as a daemon
hermes gateway status    # see which platforms are currently active
hermes gateway stop      # stop the daemon cleanly

Security Model: Allowlist by Default#

This is the rule to internalise before you turn any bot on: by default, the gateway denies every user who isn't explicitly on an allowlist or paired through DM. A bot with terminal access on a public group chat is a disaster waiting to happen, so Hermes makes 'locked down' the default state.

You have three mechanisms:

• Per-platform env vars — e.g. TELEGRAM_ALLOWED_USERS=123456789,987654321 for Telegram.
• Cross-platform — GATEWAY_ALLOWED_USERS applies the same list everywhere.
• DM pairing codes — the agent hands out one-time codes you approve from the CLI with hermes pairing approve. Good for teammates who don't want to share user IDs up front.

Never set GATEWAY_ALLOW_ALL_USERS=true on a bot with terminal tools enabled. The docs put this in bold for a reason.

Platform Capability Matrix#

Not every platform supports every feature. A rough orientation:

The table is a guide, not a contract — double-check the upstream page for the platform you're about to configure because feature coverage evolves.

Per-Platform Toolsets#

Each platform has a matching toolset — hermes-telegram, hermes-discord, hermes-slack, and so on — that determines which tools the agent can use when a message arrives from that platform. This is how you enforce 'Slack bot can read files but can't run rm' without writing custom code.

Home Assistant is the outlier worth mentioning: rather than being a messaging platform at all, it's a smart-home integration that exposes device state queries and service controls as tools. Handy if you want Hermes to flip your lights or read a sensor, less useful as a conversation channel.

Universal Chat Commands#

Regardless of platform, the same handful of chat commands work everywhere — they're the cross-platform equivalent of the CLI's slash commands:

• /new — start a fresh conversation, discarding prior context.
• /model — swap the model (subject to allowlist).
• /voice — toggle voice mode where supported.
• /background — spawn an async task, handy on mobile where you can't leave a long operation running in the foreground.

Sessions persist across messages with configurable reset policies — daily schedule, idle timeout, or on-demand only.

Recommended First Platform#

If you have no strong preference, start with Telegram. The bot model is friendly, voice transcription works out of the box, and the allowlist UX is the most self-explanatory. Once you've run Telegram successfully for a week, add the second platform — not before.

Next Steps#

Lesson 5 on integrations covers the other half of 'talking to external systems' — LLM providers, MCP tool servers, web search and voice backends. If your goal is to deploy a messaging bot for a UK business without burning a weekend on allowlist configuration, SetupClaw's managed service handles the gateway hardening end-to-end.