Set it up

Install, pick a model and connect three platforms. Then spend the rest of the evening on access control.

1 · Install and pick a model

curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

iex (irm https://hermes-agent.nousresearch.com/install.ps1)

Run hermes once and connect a provider: Nous Portal, OpenRouter, OpenAI or your own endpoint. Start with a mid-tier model, since every message is a model call, and move up once you’ve seen the volume. Config is in ~/.hermes/config.yaml, secrets in ~/.hermes/.env. CLI args beat config.yaml, which beats .env, which beats defaults.

2 · Connect the three platforms

You need a Discord bot token, a Slack app and a WhatsApp link. Variable names and scopes change, so take them from each platform’s page in the messaging docs. Tokens go in ~/.hermes/.env, settings in that platform’s section of config.yaml. Then start the gateway:

hermes gateway

When it works in the foreground, install it as a systemd (Linux) or launchd (macOS) service so it survives reboots. Add Discord first, since it’s easiest to verify, then Slack, then WhatsApp.

3 · Lock it down before you invite it anywhere

Access is in or out: anyone who passes can command the agent. The gateway checks in this order: per-platform allow-all → DM-pairing approved list → platform allowlists → global allowlist → global allow-all → deny. Two rules:

• Never enable allow-all where strangers can message the bot. On a public Discord, every member, present and future, could command it. It’s first in the check order, so it overrides every list below it.
• Use DM pairing for people and allowlists for known IDs. An unknown user who DMs the bot gets a one-time code, gives it to you, and you approve it:

hermes pairing approve discord <code>

Pair yourself and your mods. Everyone else is community. There are no role tiers inside one instance. If mods need a different bot from the one you run, make a separate profile (hermes profile create mods, run with hermes --profile mods) with its own config, allowlist and toolsets. hermes gateway list shows gateways per profile. Most communities don’t need this on day one.

4 · Set the approval gates

Two settings in config.yaml:

• skills.write_approval gates the agent editing its own skills with skill_manage. Keep it on. The prompts have the agent write three skills, and you want to see each diff, because skill files persist and run scripts.
• approvals.mode is smart (the model judges risk), manual or off. Use manual for the first week while you see what the bot tries to do. approvals.deny takes fnmatch globs for commands that must never run, whoever asks.

5 · Cron

Jobs live in ~/.hermes/cron/, schedules can be plain language, and output can go to any connected platform (that’s how the pulse reaches your DMs). You’ll create the jobs in the prompts. These manage them:

hermes cron list
hermes cron status
hermes cron run <job>      # fire one now to test it
hermes cron pause <job>
hermes cron remove <job>

Before building a changelog fetcher, try hermes skills search and hermes skills install. They pull published skills from the Skills Hub and scan them on install.