Find your first step
Choose English or Spanish. Route 1 takes you into Desktop and explains what you need before installing.
ONE GUIDE DESKTOP FIRST EN + ES
Choose a model. Give it a task. Check the result. Then build bots, memory and routines you can direct from the same application.
Switch models to fit the work without rebuilding your system each time.
Launch: October 7, 2026 · One-time payment · Independent publication

FROM DOWNLOAD TO RESULT
The ZIP includes the reader, reference PDFs and files for each route. Extract the folder and open index.html. You can also read it from a USB drive without setting up a server.
Choose English or Spanish. Route 1 takes you into Desktop and explains what you need before installing.
Connect a provider you already use. The guide distinguishes account login, API access and local models.
Try the Faro exercise. See what Hermes delivered, where to review it and what to do if the first attempt fails.
HERMES DESKTOP AT THE CENTRE
Talk with your bots inside Desktop, inspect their individual work and switch models when the task calls for it. The guide shows you how to build and supervise that journey.
If your hardware supports local models, Hermes helps you see what fits. Otherwise you can still use Desktop, providers, voice, tools and automation.
HERMES AGENT / TECHNICAL PROFILE
Start in Desktop with a model you already use. If your hardware can run a local model, Hermes can download and run it there too. Switch models to fit the work.
Hermes detects GPU and RAM, manages llama.cpp, estimates which models fit and selects an appropriate build. You can also search Hugging Face models while Hermes manages memory and context as you work.
In Desktop, each bot can have its own model, memory, skills and chat. Review individual work, bring bots into a group and delegate tasks to subagents.
Memory, voice, browser, tools, MCP, plugins, connectors, cron and Kanban extend what Hermes can do. Work on your computer or run an installation on a server when you need continuity.
Local work depends on your hardware. Nous documents 8 GB VRAM for smaller models and 16 GB or more for 27–35B models. Actual speed and fit depend on the model, context and your machine. Offline work only applies when your workflow uses no remote tools or services.
The Local Models interface is enabled in Canary builds; other Desktop builds need the --local launch flag. The guide shows how to prepare, run and check this path. Read the official documentation ↗
WHAT YOU GET
Open the ZIP. Choose your language. Follow Route 1 in Desktop and produce a result you can check. Then move on to bots, models, tools and automations with prepared files for each exercise.
The guide takes you through concrete tasks. The guide includes an exercise for finding and assessing Hugging Face models, plus an index for locating plugins and connectors. You configure external accounts and integrations with your own permissions. Downloading software, connecting providers and using cloud services require Internet access and sometimes an account or API key.
BEFORE YOU BEGIN
No. Route 1 starts in Desktop with a provider. Local models are optional and Hermes helps assess what your computer can run.
No. You get an offline reader, copyable and downloadable files, and two reference PDFs. Each practical route explains the expected result and what to check if it fails.
The reader and materials open without a server. Installing Hermes, connecting accounts and downloading models still require their own steps. Cloud models require connectivity.
No. It is an independent publication by Cosas Agénticas. Get Hermes Agent from the official project.
ONE GUIDE. THE COMPLETE JOURNEY
An English and Spanish edition to learn, test and return to when you need it.
Founding price · €9
From October 7 to October 18, 2026 at 00:01 Madrid time. 11 days at €9. Then €29.
Buy the guide for €9Secure checkout with Stripe. You will receive an email link to download the ES/EN ZIP. If checkout does not open, email us.