Loading

← Hermes Agent Operator Track

Messaging gateway — put Hermes where you live

> One gateway process: Telegram, Discord, Slack, WhatsApp, Feishu, Teams, and more.

Canonical: https://www.brainyxai.co.za/education/hermes/messaging-gateway

Markdown: https://www.brainyxai.co.za/md/education/hermes/messaging-gateway.md

Course: Hermes Agent Operator Track

Lesson: 7 of 9

Minutes: 18

Author: Brainyx AI

What you will be able to do

  • Explain the gateway as a single multi-platform process
  • Pick a first platform for a personal bot
  • List prerequisites before going always-on

One gateway, many surfaces

The messaging gateway is a background process that connects configured platforms, manages sessions, runs cron, and can deliver voice. Platforms documented include Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Mattermost, Email, SMS, DingTalk, Feishu/Lark, WeCom, Weixin, QQ, Microsoft Teams, Google Chat, Home Assistant, and more — see the [Messaging Gateway](https://hermes-agent.nousresearch.com/docs/user-guide/messaging) comparison table for feature support (voice, images, threads, streaming).

Prerequisites before always-on

1. Working local chat with a paid/credited model path

2. Tools you need (search/TTS/browser if the bot should use them) — Portal bundles help

3. Authorization — allowlists / DM pairing (next lesson) so strangers can’t drive your shell

4. A host that stays up — laptop sleep kills bots; prefer a small VPS

Voice

Voice mode spans CLI mic, messaging spoken replies, and Discord VC. Follow [Voice Mode](https://hermes-agent.nousresearch.com/docs/user-guide/features/voice-mode) and the [voice guide](https://hermes-agent.nousresearch.com/docs/guides/use-voice-mode-with-hermes) after text chat is solid.

Cron & delivery

Scheduled automations can deliver to any configured platform ([Cron](https://hermes-agent.nousresearch.com/docs/user-guide/features/cron)). Start with one daily digest job before building a fleet of schedules.

Brainyx AI production note

For client WhatsApp/Telegram agents we still design owned workflows (CRM hooks, approval, audit). Hermes is an excellent personal/ops control plane and a reference architecture — enterprise messaging often still needs custom integration layers.

Gateway lab

1. Choose one platform (Telegram or Discord is a common first).

2. Follow the official platform page under Messaging Platforms to create a bot token — store it only in Hermes secrets/config, never in a public repo.

3. Send yourself a message and confirm the gateway session is the one you expect.

Checkpoints

  • I understand the gateway is always-on infrastructure
  • I will not expose an unauthorized bot to the public internet

Crossing from private to reachable

The gateway is the point where a personal tool becomes something other people can talk to. Everything about your risk posture changes at that line, and the controls need to exist before you cross it, not after the first surprise.

Three are non-negotiable. Allowlisting — explicitly named people or channels, never open to anyone who finds the bot. Rate limits — per user, so one person or one loop cannot exhaust your budget. Logging — who asked what, what tools ran, what was returned.

Without these you have published an endpoint that spends your money and acts in your systems on behalf of strangers.

Picking a first platform

Choose based on where the conversation already happens, not on which integration looks most interesting.

For a personal assistant, use whatever you already have open all day. For an internal team bot, use your team's existing channel — adoption of a bot in a tool nobody opens is zero regardless of quality. Only move to customer-facing platforms once the internal version has run without incident for a while.

Start with one platform. Multi-platform is a configuration exercise; the hard part is the operational discipline, and that is easier to learn on one surface.

Going always-on

An always-on bot is a service, with the obligations that implies: it must restart after a crash, survive a reboot, log somewhere you will actually look, and have an owner who notices when it is down.

Local hardware and load-shedding are a poor combination for anything people depend on. If uptime matters, host it somewhere with stable power and connectivity, and decide in advance whether silent downtime is acceptable — if not, you need monitoring, not hope.

WhatsApp and POPIA in South Africa

WhatsApp is where most South African customer conversation happens, which makes it the obvious target and the highest-consequence one.

Two things follow. Customer-facing WhatsApp is a business platform with its own template, opt-in, and approval rules — not a personal-bot integration. And every message is personal information: conversation logs need a retention policy, access control, and a lawful basis, and if the agent can act on customer accounts, an audit trail is mandatory.

For production WhatsApp support with escalation and compliance handled properly, that is [AI chatbot development](https://www.brainyxai.co.za/services/ai-chatbot-development) rather than a gateway configuration.

Mini-FAQ

A: Yes, one gateway process handles multiple. Get the controls right on one first.

A: It should refuse and log the attempt. Verify that behaviour deliberately before you share the address.

A: Require an explicit mention or command prefix. Agents that respond to every message in a channel get muted within a day.

Next lesson

Continue to [MCP integrations](https://www.brainyxai.co.za/education/hermes/mcp-integrations).

Official reference: https://hermes-agent.nousresearch.com/docs/user-guide/messaging

Course hub: https://www.brainyxai.co.za/education/hermes · Previous: https://www.brainyxai.co.za/education/hermes/skills-and-memory · Next: https://www.brainyxai.co.za/education/hermes/mcp-integrations

Return to the course hub, or explore Brainyx AI services to put the skill into a live system. Markdown: /md/education/hermes/messaging-gateway.md

Book a consultation · joshua@brainyxai.co.za · Markdown mirrors