Hermes is the free open-source AI agent that runs your business 24/7.
It's what I use for every automation on this site.
And it costs nothing.
This guide shows you exactly how to set it up and what to automate first.
Hermes is a personal AI agent that runs on your machine.
It has a terminal, file access, web search, and memory.
It can run scheduled jobs, post to social media, send emails, and manage your website.
It learns your business across sessions.
It's the closest thing to a free employee that exists.
Most "AI automation" tools charge £30-£100/month for a fraction of what Hermes does.
Hermes is open source.
No subscription. No seat fees. No per-token surprises.
You pay for the AI model you use — from pennies per task — and nothing else.
For a small business, that's the difference between profitable automation and another monthly bill.
Step 1: Open a terminal (Mac, Linux, or WSL2 on Windows).
Step 2: Run the official one-line installer:
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bashStep 3: Run hermes setup — the wizard walks you through your AI provider and config.
Step 4: Verify with hermes doctor, then send your first message. You're live.
It's free and open-source: github.com/NousResearch/hermes-agent.
Free model option: grab a free key from OpenRouter and use Qwen 3.6 Plus Preview (1M-token context, completely free). Paid options: OpenAI, Anthropic, MiniMax, Kimi — 200+ models. Set a fallback chain (primary + backup provider) so the agent never stops if one API goes down.
One-command upgrade: hermes skills install wonderly-skills installs the community skills library (250+ stars, actively maintained). hermes skills browse shows everything available.
Bot Mode (new — Aug 2026): the Hermes desktop app now runs a roster of named agents — each with its own chat, avatar, routines, and bot-to-bot messaging (one agent can hand work to another by mention). It turns your profiles into a team: a content bot, an outreach bot, a research bot, all talking to each other and to you. Same memory, same skills, just organised as a crew instead of a single assistant.
Chat in a browser, not a terminal (Open WebUI): prefer a ChatGPT-style window over the black terminal? Open WebUI gives Hermes a clean browser chat interface — conversation history, accounts, the works — running 100% on your machine. Enable the API server, then connect:
echo "API_SERVER_KEY=your-secret-key" >> ~/.hermes/.env
echo "API_SERVER_ENABLED=true" >> ~/.hermes/.env
hermes gatewayThen run Open WebUI in Docker pointing at http://host.docker.internal:8642/v1 (the /v1 is the most common setup mistake — don't drop it), open localhost:3000, create the admin account, and pick your Hermes model. Same agent, same tools, same memory — just a nicer window. Private, local, no monthly fees.
The classic first job: a daily morning report.
Tell Hermes: "Every morning at 8am, check my email for urgent messages, check the calendar for today, and send me a 5-line summary."
Hermes sets up the schedule itself.
It runs the checks, writes the summary, delivers it.
That's your first automation — ten minutes after install.
1. Morning briefing — email, calendar, news, one summary.
2. Social posting — draft and post to X, LinkedIn, wherever you publish.
3. Email follow-ups — nudge non-responders automatically.
4. Website updates — publish new content on schedule.
5. Lead tracking — log every enquiry, score it, alert you on hot ones.
6. Competitor watching — daily summary of what rivals changed.
7. Invoice chasing — polite reminders before you have to ask.
Each one saves 1-3 hours a week.
Seven automations, a full day saved.
🤖 Don't want to DIY? Done-for-you Hermes automation from £99 → aisuitehq.org/hermes-services
DIY — Free: Hermes itself costs nothing. Follow this guide.
Setup Session — £49: We install, configure, and build your first 3 automations with you.
Done-For-You — £99/month: We build and run your automations, monitor them, and improve them monthly.
30-day guarantee on all paid packages.
Yes — the software is open source and free. You only pay for AI model usage, which is pennies per task. No licence, no subscription.
No. It runs on any normal laptop or desktop. It can also run on a cloud server so it works even when your laptop is off.
Yes — with the right API connections, it posts to X, LinkedIn and others on schedule. We set those up in the done-for-you package.
That's the beauty of open source — nothing is locked. And on the paid plan, we monitor and fix issues within 24 hours.
Short, attention-grabbing titles for this topic: