Hermes Agent for business works best as a small team of AI assistants, one per role, that live in your chat apps, run on a schedule and ask before doing anything risky. Hermes Agent is the open-source agent from Nous Research. It runs on your own computer or server, connects to the model provider you choose, and reaches your team through Telegram, Slack, WhatsApp, Email, Teams and more.
📺 Watch: New Hermes Agent OS Just Changes EVERY THING!
🔥 Get the Agent OS as a free bonus: AI Profit Boardroom members get the full Agent OS zip, prompt libraries, daily tutorials and weekly live coaching calls. → Get inside
This guide shows how a small business would set it up: which features to use, in what order, and how to keep it safe and affordable.
Why Hermes Agent for business is different from a chatbot
A chatbot answers a question and forgets. Hermes keeps working. It schedules its own jobs, remembers what it learned, loads skills when a task needs them, and hands sub-tasks to helper agents. The same agent core runs in the terminal, in the Hermes Desktop app for macOS, Windows and Linux, in a browser Webapp, and in your team chat. Sessions are shared, so a task started on Desktop can be picked up from another front end.
The latest release is v0.22.7, a patch release published on 3 November 2026. The feature release before it, v0.22.6 (2 November 2026), added a Stop button in the Desktop cron editor, a per-job missed-run policy for cron, every profile shown as its own Desktop sidebar section, and a hook that can block a turn before the model is called.
Want the exact business setups other owners are running? Join the AI Profit Boardroom ($69/mo) for templates, walkthroughs and weekly calls. Or book a free AI strategy session for a plan built around your business.
The Hermes Agent for business building blocks
| Feature | What it does | Business use |
|---|---|---|
| Profiles | Separate Hermes homes, each with its own config, keys, memory, skills, sessions and cron jobs | One assistant per role: sales, support, ops, content |
| Messaging gateway | Connects Telegram, Discord, Slack, WhatsApp, Signal, Email, Teams and more | The team talks to the agent where they already work |
| Cron jobs | Natural-language or cron schedules, with skills attached and results delivered to chat or files | Morning briefs, weekly reports, recurring checks |
| Skills | On-demand instruction packs loaded only when needed | Your processes written down once and reused |
| Memory | Bounded MEMORY.md and USER.md that persist across sessions | The agent keeps your preferences and conventions |
| Delegation | Subagents with isolated context; only a summary returns | Parallel research and long tasks without clutter |
| Approvals and hooks | Smart, manual or off approval modes, deny rules, shell hooks that can block | Guardrails before anything risky runs |
| Desktop and Webapp | Native app and browser access to the same agent | Non-technical staff use it without a terminal |
Set up Hermes profiles per business role
Profiles are the foundation. Each profile is its own Hermes home with its own config, API keys, SOUL.md personality file, memory, sessions, skills and cron jobs. Create a profile called sales and you get a sales command straight away, so you can run sales chat or start its gateway on its own.
Give every role its own profile. The Hermes docs warn never to point two agent processes at the same profile, because both write memory and each loads the other's notes. Separate profiles keep the support assistant's memory clean of sales notes, and the reverse. Our guide to running multiple Hermes profiles covers the commands.
📺 Watch: Run AI Agent 24/7 on Your Hermes AgentOS
Connect the Hermes messaging gateway
Run hermes gateway setup once the CLI works. Pick the platform your team uses, such as Telegram or Slack. Hermes then answers in that chat with the same tools and memory as the terminal. Its security model starts with user authorization, using allowlists and DM pairing, so only approved people can talk to the bot. If you are choosing a platform, see the best messaging app for Hermes Agent.
Automate business routines with Hermes cron
Cron jobs turn Hermes from an assistant you ask into one that reports to you. A job can run once or on repeat, attach one or more skills, and deliver the result back to the chat it came from, to a local file, or to another platform target. A job can also run in no-agent mode, which runs a script on schedule with no model involved at all. That is the cheapest way to handle simple checks.
The Hermes Agent cron job guide walks through creating, pausing, editing and cancelling jobs.
📺 Watch: NEW Hermes Project Manager Agent is WILD!
Hermes skills for your processes, memory for the details
Skills are instruction documents the agent loads only when a task calls for them, which keeps token use down. Write one for each repeatable process: how you qualify a lead, how you format a client report, how you reply to a refund request. Hermes follows the agentskills.io open standard, and v0.22.6 added per-skill locks so a skill you approved cannot be rewritten by the agent on its own.
Memory is deliberately small. MEMORY.md holds environment facts and conventions, and USER.md holds preferences and communication style. Keep big documents in skills or files, and let memory carry the short rules the agent should always know.
Delegate big jobs to Hermes subagents
The delegate_task tool spawns child agents, each with a fresh conversation and its own terminal. Only the final summary returns to the main chat, so a long research job does not flood the conversation. You can set a separate model for subagents through the delegation settings. Read more in Hermes delegation.
Give staff Hermes Desktop and the Webapp
Not everyone wants a terminal. Hermes Desktop is a native app built on the same agent, config, skills and memory as the CLI. The Webapp, opened with hermes webapp, serves the agent in a browser. If you open it beyond the local machine, Hermes requires a dashboard sign-in provider. See the Hermes Webapp guide for setup.
Hermes guardrails for business: approvals and hooks
The default approval mode is smart. An auxiliary model reviews flagged commands, approves low-risk ones, denies clearly risky ones and escalates the rest to you. Manual mode asks before every flagged command, and on messaging platforms it queues a pending approval. Deny rules block matching commands outright, even when approvals are switched off. For business use, start on manual for any profile that touches client data, then move to smart once you trust the setup.
Hooks add your own checks. A shell pre_tool_call hook can block a tool call or fail closed, and outbound webhooks can push signed events to your dashboards. The Hermes Agent hooks guide has examples. Not sure which guardrails your setup needs? Talk it through in a free AI strategy session.
Control Hermes Agent costs with model choice
Hermes does not lock you into one model, and that is where cost control lives. Pick a strong model for the main chat. Pin a lighter model to each routine cron job; a job checks its own pin first, then a cron-wide default, then your main model. Set a separate model for subagents. Side tasks such as titles, vision and compression run on auxiliary models you can override per task. Use no-agent cron jobs for checks that need no model. For picks, see the best Hermes Agent models. For Julian's own model testing, see Goldie Bench.
Four illustrated Hermes Agent business scenarios
These show how a business could use Hermes. They are illustrations, not customer stories.
A five-person agency
A five-person agency could run three profiles: ops, content and sales. Ops posts a Monday brief to Slack from a cron job. Content drafts posts from a skill that holds the brand voice. Sales researches inbound leads through subagents and drops a summary in the sales channel.
A local service business
A plumbing or cleaning firm could connect one profile to WhatsApp for the owner. It could summarise new enquiries each evening and draft replies, with manual approval on anything that sends a message.
An online store
A store owner could schedule a nightly job that reads the day's support emails, groups common questions and writes a short report to a file, using a lighter model pinned to that job.
A consultant working solo
A consultant could use Hermes Desktop for client research during the day and a Telegram bot on the road, both on the same profile and memory.
For more ideas, see the best Hermes Agent use cases and the best AI agents for small business. For the full system view, read the Agent OS guide. Members of the AI Profit Boardroom get step-by-step builds of setups like these.
Hermes Agent for business FAQ
Is Hermes Agent safe for client work?
Hermes has layered security: user allowlists on messaging, command approvals, deny rules, file write safety and optional container isolation. Use manual approvals on profiles that touch client data.
Do my staff need to use a terminal?
No. They can use Hermes Desktop, the Webapp or a chat bot in Telegram, Slack or Teams. Setup itself is easier in the terminal.
Can each team member have their own assistant?
Yes. Create a profile per person or per role. Each one keeps its own memory, skills, keys and cron jobs.
Where do I start?
Start with one profile, one chat platform and one cron job that saves you time every week. Then add skills. Want a plan for your business? Book a free AI strategy session.
