// install · hermes
Hermes
The same files as OpenClaw (identity, memory, skills), but in a worker's casing: the agent wakes itself up on crons and heartbeats and runs the tasks without waiting for you to ask. Set up on your side, 7–12 days, flat 1 500–2 500 PLN.
// what hermes is
Hermes is not a "Slack bot". It's a work environment for an agent. It gets a procedure, works on files, terminal, web and APIs, saves the result, uses skills, runs asynchronously. Underneath it has the same file structure as OpenClaw: who it is, what it remembers, what it can do. The difference: heartbeats and crons. The agent wakes itself up, follows the task scheduled for that hour, checks what changed, delegates subtasks, leaves a work log. It doesn't wait for you to ask.
// what it's built from
What it's built from.
An agent isn't magic or "a model". It's the sum of its files. It starts every session by loading them, so it stays continuous even though it technically restarts from zero each time. These are its organs.
Identity: who it should be
Here we define who the agent is: tone, character, boundaries. For many that's a detail, it can simply be plain and impersonal. For some it isn't: my wife likes that her assistant is Dwight, who says "FACT" before answering a question. It can be your favourite fictional character. Identity also covers the model choice: the agent can run on a local LLM or on the best one available. Are you a fan of Gemini, Claude, or Codex? Any of them can be your agent. It's one file, you edit it yourself.
Memory: what it remembers
Default AI forgets everything after the session. Here it's different. The agent has a file structure it loads at the start of every conversation: company context, decisions, your preferences. And it grows it itself. Over time it records what you like about its work, what not to do, when to use which skill. After a week it knows more than at the start.
Tools and skills: what it can do
What the agent can actually execute: concrete procedures, integrations, actions. Each skill is a file with a prompt and guardrails. The agent knows when to run it and when to stop and ask a human. We add and refine skills as the scope of its work grows.
Rules: how it behaves
The system prompt: how it should act, what it never does, when it escalates to you instead of guessing. Your marketing team can review the tone, legal can review the guardrails. You edit it yourself, no ticket.
You change a file, and the agent works the new way from the next conversation. No training, no ticket. The speed of a change is the speed of editing a text file. More on how we build memory: why AI forgets and how to fix it.
// how hermes uses this
Hermes takes those organs and adds a schedule: a cron ("do this every day at 7:00") and a heartbeat ("check every hour whether something changed"). So the agent isn't reactive: it has a plan for every hour and executes it. It holds context for a long time, has proper memory, delegates subtasks. From the client's side: you describe a procedure once, we set the schedule, and the result simply shows up where it belongs: a file, a report, Slack, email.
// what it actually does for you
What you can hand it.
Hermes shines where a procedure has a deadline and nobody feels like doing it. A few examples, to show the range:
Weekly competitor research
A cron: every Monday morning the agent searches competitors' sites, social media and listings, picks the changes per your criteria, assembles a report. You read a summary instead of visiting ten sites.
A scan of the creators you follow
Favourite people on YouTube, X, Threads, Instagram? A cron wakes the agent at a given hour, scans the specified people and portals, picks material per your goals, and either gives a brief/summary or, if something is more important, leaves the raw material. Hours of reading and listening disappear.
A daily brief: politics, the economy, sports, company finances
Another cron: every morning the agent gathers and summarizes what's happening in the topics that concern you, or a financial roundup of the companies on your list. One place instead of fifteen tabs.
Triage of incoming leads or emails
An email/lead comes in → the agent reads, classifies, prepares a draft reply or a system entry, routes it to the right person. A human reviews and approves instead of reading everything from scratch.
Monitoring changes you have to catch
A heartbeat: the agent checks every hour/day whether a status, price, document or availability changed, and lets you know only when something actually changed.
The common denominator: "we have a procedure, but nobody runs it consistently" or "I want to get X at hour Y without asking". Anything you can describe that way is for Hermes.
// who it's for
Who it's for.
You have SOPs, checklists, weekly research, monitoring. None of it gets done consistently, because it's dull, repetitive work.
You're not looking for a chat to ask questions. You want something that runs a task end to end and leaves the result where it belongs.
Daily lead review, weekly research, change monitoring: things that have a deadline and nobody feels like doing.
// what you get with the install
What you get with the install.
Discovery 60–90 min
We map 2–3 procedures to automate: which steps, what input, where the result should land, who approves.
The first version of its files
Identity, memory with company context, 2–3 procedures described as skills (workflow definitions, prompts, guardrails), rules. First polish from me.
Custom Hermes setup for your stack
Models, providers, tools, all configured for what you use. BYO model (Anthropic, OpenAI, Gemini, local).
Crons and heartbeats per procedure
When it runs, how often, what it checks. The agent wakes itself up and follows the plan.
Logs and monitoring
You see what the agent did, when, and with what result. Not a black box. A full work trail.
3–7 days of tuning after install
First runs → review → skill fixes. The procedures settle in on real data.
"How to improve the workflow" docs
How to change a step, add a procedure, set a new schedule. You extend it further yourself or with the team.
// how it looks
How it looks.
We take a real procedure
E.g. weekly research or lead triage: something a human does today, or nobody does.
I assemble the agent's first files
Identity, memory, 2–3 procedures as skills, rules, schedules. Configured for your stack.
The agent runs it on real data
Not a demo on made-up data. Your sources, your context, a real result where it belongs.
We fix the skills, the next run is better
Feedback after the first run → workflow definition fix → next run with fewer manual steps. From there the agent learns from your feedback.
// what you provide
What you provide.
I can do the install without your involvement: a default setup for a typical procedure in your industry. But the more precisely you describe how the procedure really looks and what not to do, the fewer fixes later. We do the custom together at the start or in the first update.
- Your time for discovery. I need to understand the procedures you want to automate and pull the context from you. I handle the rest of the technical side.
- A ChatGPT Plus subscription with OAuth ($20/mo) or an Anthropic API key with a budget.
- A local or VPS environment (Mac, Linux, Docker).
- Access to the data sources the procedures need (repo, Notion, email, API: whatever the procedure touches).
// scope
Scope.
- The first version of the agent's files (identity, memory, 2–3 procedures as skills, rules)
- Custom Hermes setup for your stack (BYO model)
- Crons / heartbeats per procedure
- Logs + monitoring + handoff for the team
- 3–7 days of tuning + documentation
- Multi-agent kanban orchestration (advanced; separate sprint)
- Integrations with legacy systems (ERP, enterprise CRM)
- Voice/mobile UX (that's OpenClaw territory)
- RL training / trajectory tooling
// timeline · effort · price
Timeline, effort, price.
7–12 working days from discovery to a working install
~12 hours of my work per install
1 500–2 500 PLN flat; the price is fixed in this phase (first-wave promo; reverts to 3 500–4 500 after the first case-study clients)
Optional 1 000 PLN/mo, first month free. Most clients of this type keep the retainer; models rotate, keys expire, the workflow wants tuning.
The LLM subscription or API key is on your side (ChatGPT Plus with OAuth ~$20/mo preferred: flat, no per-query burn; or an Anthropic/OpenAI/Gemini key). Contract disclaimer: token cost is yours.
// video
A walkthrough and demo of how it works
Coming soon: a clip where I walk through the agent's files and show Hermes running a real procedure on a cron. For now the fastest path is a free diagnostic call.
// faq
FAQ
How is Hermes different from n8n or Make/Zapier?
n8n and Zapier are "if X, do Y": fixed paths. Hermes gets a goal and decides for itself which steps to take and how to handle surprises. Plus it holds context for a long time, has proper memory, and wakes itself up on crons. It's a worker agent, not an if-statement automaton.
What are a cron and a heartbeat?
A cron is a schedule: "run this every day at 7:00", "on Mondays prepare a report". A heartbeat is a regular "check whether something changed", every hour or every day. Thanks to them the agent works on its own, without your command.
How do I know the agent did it right?
Every run leaves a log: what it took as input, which steps it ran, what result it produced. Plus guardrails in the skill say when the agent should stop and ask a human instead of acting blind. The first weeks are tuning that threshold.
What if the procedure changes?
You edit the workflow definition: it's a text file, not code. From the next run the agent works the new way. The "how to improve the workflow" docs are included.
How is it different from OpenClaw?
The same files (identity, memory, skills), a different casing. OpenClaw lives in the team's channels: you message, it answers. Hermes is a background worker: it runs procedures itself on crons. Together = the Hybrid Agent Stack.
What if I order the install without my involvement?
You can: a default setup for a typical procedure in your industry. It'll work, but the custom (your real procedures, your criteria) we do together at the start or in the first update. That's when it makes sense.
Do I pay for tokens separately?
Yes. Your own LLM subscription or API key. Hermes supports many providers (Anthropic, OpenAI, Gemini, local), so you pick by cost and privacy. Token cost is on your side.
// start
Where can Hermes help you?
That's where we start. Free diagnostic call, 30 min. We map 1–2 procedures and I tell you straight whether Hermes makes sense here.