AgentsUpdated October 2026

The Hermes Playbook

Your own chief of staff. Open source. Run it where you choose.

Hermes is an open-source AI agent you can run on your own machine or server. This playbook covers the current v0.21.5 release, local models, profiles, Bot Screen, messaging, subscription and API options, and the automations that run Ential and the Drake household.

Start reading

Ryan Drake

Founder, Ential · Updated October 2026 · 8 min read

01. What Is Hermes?

An open-source AI agent that lives on your machine and actually does the work.

Every AI tool you use in a temporary thread starts with a context problem. Hermes is built around persistence. It is an open-source agent that runs on your machine or server, takes real action through tools, and reports back through the terminal, desktop or a messaging gateway spanning 21+ platforms when native and plugin channels are counted together.

The difference from a normal AI assistant is agency. A chatbot answers a question and forgets you. Hermes holds context about you, your business, and your family, then runs recurring work on a schedule without being asked. Think of it less as a tool and more as a chief of staff that never sleeps.

Everything in this playbook is a real, running Hermes setup. At Ential it shows up as Donna: the assistant that runs the morning brief, triages the pipeline, plans date night, and keeps the books honest. By the end you will have the same operating pattern on infrastructure you choose.

"Your agent. Your machine. Your rules. Your choice of model provider."

Runs 24/7 on your own computer or a server
Connects to email, calendar, CRM, project tools, and messaging apps
Runs recurring jobs on a schedule (briefs, reviews, reminders)
Supports subscriptions, API providers and local open models
Self-hosted runtime. Model and connector choices still determine where data goes
Extensible with skills for anything you do more than once

Official documentation

02. Why It Matters

Most people top out at copy-pasting into a chat window. It helps, but nothing compounds: you still start every task, still remember every follow-up, still do the work of asking. The leverage only shows up when the agent runs on its own and the results stack week over week.

The usual objection is cost. A busy metered API can run into hundreds of dollars a month, while a subscription has usage rules and a local model has hardware costs. Hermes lets you choose and route between those models instead of locking the whole agent to one commercial plan.

Unlock the Other 9 Chapters

Free with your email. Each new playbook lands in your inbox the day it ships.

02Why It Matters
03Hermes vs Claude Cowork
04What The Latest Release Added
05How To Get Started
06The Setup
07The Model Strategy
08Your Business On Autopilot
09Your Personal Life On Autopilot
10Have It Installed For You

Choose The Right Agent

Start with the job, the interface you will actually use, and the data boundary you can defend. Prices and access below are current as at 4 October 2026.

Agent	Best for	Interface	Runs where	Autonomy	Access and price	Data and control
Grok Bot	Cloud teams and software work	Bot app, desktop and connected tools	A shared cloud computer	Always on, specialist bots can hand work off	Cursor Pro from US$20/mo or SuperGrok from US$30/mo	Vendor-hosted, bots share the computer and its signed-in sessions
Instinct	Personal errands, shopping and persistent chases	Text, phone call and one continuing conversation	Instinct-managed phones and computers	Persistent personal operator with proactive follow-up	Private access, no public first-party price as at 4 Oct 2026	Vendor-hosted, broad optional access to personal apps and device context
HermesReading now	A custom agent you can operate and extend yourself	Terminal, desktop and 21+ messaging platforms	Your Mac, Linux machine, server or supported cloud host	Schedules, subagents, skills, memory and multi-profile teams	Open source, model, subscription and hosting costs are separate	Self-hosted by default, control depends on the model and connectors you choose
ChatGPT Dots	Ongoing knowledge work inside the ChatGPT ecosystem	ChatGPT, Slack, Teams, voice and limited SMS beta	OpenAI cloud computer plus connected environments	Always-on primary dot that learns from feedback and delegates work	First dot included with eligible Pro or Business Premium plans	OpenAI-hosted, permissions and approval boundaries require active review
Meta Muse	Personal assistance and small-business back-office work	Muse app, Mac and connected business services	Meta cloud systems with optional Mac connectors	Background goals, calls, browsing and connector-driven tasks	Free limited tier, Power US$20/mo, Maximum US$100/mo	Meta-hosted, connector permissions are optional and revocable

Swipe the table sideways to compare every column.

Skip the build

Have Donna Installed In Your Life And Business

Donna is this exact Hermes setup, done for you. We install and tune your models, your channels, your business tools, and the personal automations that matter, around how you actually work, until it just runs.

Your models and channels wired up and signed in
Business tools connected: inbox, calendar, CRM, the books
Personal automations set: briefs, reviews, date night, reminders
Tuned until it just works, then it runs on autopilot

$3,990 to set up, plus $99 a month to keep her running. An executive assistant costs $85,000 a year. Donna’s first year costs roughly 4 weeks of one, and she never resigns.

More Playbooks

Agents

The Grok Bot Playbook

Read playbook

Agents

The Instinct Playbook

Read playbook

Agents

The ChatGPT Dots Playbook

Read playbook