Mykleos is an assistant that lives on a mini-computer at home,
talks through Telegram and from the terminal,
and can do things on your behalf asking permission when it matters.
Think of a butler in the kitchen, twenty-four hours a day. He doesn't
leave the house without asking, doesn't spend money without warning you,
doesn't delete anything he hasn't shown you first. But if you say "download
yesterday's attachment and put it in Documents", he does it, and
reports back: "done".
It is not a chatbot. A chatbot answers questions. Mykleos
acts: downloads files, reads email, takes notes, checks the
calendar, sums up conversations. It answers questions too, of course, but
its reason for being is to do small things on your behalf.
2. Six scenes from real life
The quickest way to understand what it does is to walk through a few
examples. The first three are reactive (you ask, it answers); the other
three show the more interesting side: when it acts on its own at set times,
when it delegates to another home agent, when it builds itself a new tool.
Scene 1
"Summarise today's emails"
Roberto opens Telegram from the train. He writes to his Mykleos bot:
“Hi, give me a summary of today's important emails”
Mykleos reads the inbox, drops the spam, groups the mails by sender and
by priority, and replies in 4-5 lines: "3 emails today: (1) the lawyer
asks for the document by Friday, (2) the bank confirms the wire transfer,
(3) your sister about Sunday's lunch. Shall I reply to any of these?"
No approval needed: reading email is considered "safe"
(Mykleos doesn't modify anything, doesn't send anything).
Scene 2
"Reply to the lawyer for me"
Roberto, still from the train:
“OK, you reply to the lawyer, tell him I'll send the document by Thursday”
Mykleos composes the draft, and before sending it asks for
confirmation, showing the exact text:
✋ May I send? To: studio.rossi@lawfirm.example Body: Dear Sir, I confirm I will send the document by Thursday
24 April. Kind regards, Roberto.
[ Yes, send ] [ Wait, let's edit ] [ No, drop it ]
Roberto taps "Yes, send", the email goes out. Two seconds later:
"Done. Delivered at 11:34."
Scene 3
"Remember that tomorrow is Mum's birthday"
Roberto, in the kitchen, says to the terminal:
“Tomorrow is Mum's birthday. Write it down wherever it belongs.”
Mykleos updates the calendar (after approval), adds a note in
MEMORY ("Mum's birthday is on 23 April"), and the next morning
sends a Telegram message: "Today is Mum's birthday. Shall I remind you
to call her in the afternoon?"
Long-term memory: next year, on the same day, it will
already know.
Scene 4
"Every Sunday at 9, give me the weekly digest"
Roberto, one Saturday evening, writes once to Mykleos:
“From now on, every Sunday morning at 9, send me a Telegram digest of the week: important unanswered emails, deadlines in the next 7 days, notable expenses, things I had promised to do.”
Mykleos confirms: "OK, I've set a recurring reminder. First one is Sunday 26 April. Want me to try a preview now?"
The following Sunday, at 9:02, a spontaneous message arrives:
☕ Good morning. Week digest, 20–26 April
• Unanswered emails: Mr. Rossi (document by Monday), cousin Luca (christening)
• Upcoming deadlines: electricity bill 29/4, medical check-up 2/5
• Notable expenses: €312 home hardware (Wed 23)
• Promises to close: book Rome hotel, call Mum
Shall I draft replies for the two emails?
Proactive action: Mykleos woke up on its own, worked in
the background, and found you. This is possible because at the time of the
original request you gave the green light for this specific recurring
activity. Nothing arrives out of the blue without you having said yes at
least once.
Scene 5
"Ask the house if the washing machine is done"
Saturday afternoon, Roberto from the sofa:
“Ping the house, ask it to check whether the downstairs washing machine has finished its cycle, and if it has, to let me know when I'm back in the kitchen.”
Something interesting happens here: Mykleos does not control
the washing machine itself. Home automation (sensors, switches, appliances)
is the job of another agent — the home agent — which
lives on the same mini-computer and talks to the home network. Mykleos
delegates:
Mykleos → house: "status washing_machine_bathroom" house → Mykleos: "cycle ended at 2:07 PM, consumption 0.42 kWh" Mykleos → house: "notify presence_kitchen(roberto)" Mykleos → Roberto: "The washing machine finished at 2:07 PM.
I'll let you know when you're back in the kitchen."
General principle: Mykleos doesn't rewrite half your
digital home. It works with the specialised agents already there. The home
agent for voice and sensors, Mykleos for emails, files, appointments. They
speak to each other through a common protocol, like two colleagues at work.
Scene 6
"Write me a script that sorts photos by date"
Roberto has 2,400 photos downloaded from his phone in
~/Pictures/raw/, names like IMG_4812.JPG, dates
only in the EXIF metadata. He asks:
“In the Pictures/raw folder I have 2,400 photos. Put them in subfolders 2025-03, 2025-04, ... based on shooting date. Don't touch the originals: copy them.”
Mykleos realises a script is needed, writes one (Python +
exifread), and before running it, shows you:
✋ May I run? Script:~/Pictures/.myclaw/sort_by_date.py (42 lines) Source folder:~/Pictures/raw/ (read-only) Destination folder:~/Pictures/sorted/ (new, to be created) Operation: copy, does not move or delete Sandbox:supervised (no network, no other paths)
[ Yes, run ] [ Show me the script ] [ Cancel ]
Roberto clicks "Yes", the script runs inside a sandbox (can't touch
anything other than those two folders, can't reach the internet), and 90
seconds later: "Done. 2,403 photos copied into 18 subfolders, from
2023-01 to 2026-04. Originals untouched."
Why this matters: Mykleos doesn't have "a command to
sort photos". It wrote one on the spot, under sandbox supervision,
for that one request. If tomorrow you ask it to do the same on a different
folder, it can reuse the script; if the photos were in a different format,
it will write another one.
3. Where it lives and how you talk to it
Mykleos runs on a small Ubuntu computer that sits in your home (in the
living room, in the basement, under the TV — it's not big). You don't
need the internet to talk to it if you are at home; the internet is needed
only for things that require external services (Drive, email).
It has two faces. Same mind, two surfaces.
The phone face: Telegram
Figure 1 — What you see in Telegram. The style is that of a normal chat app: text, reply bubbles, buttons under confirmation requests. No commands to learn.
The home face: the terminal
Those who like the computer can talk to it from a terminal window, and
see what it is doing in real time. For those who prefer Telegram, this is
never needed.
Figure 2 — The same mind, seen from a terminal. The greyed italic lines show what Mykleos is doing "behind the scenes": useful if you want to understand, ignorable otherwise.
4. What happens when it says "may I?"
The most important piece of design, from the user's perspective. Every
time Mykleos is about to do something that changes the state of the
world (send an email, write a file, spend money), first it
stops and asks. It is not a bug: it is its number-one
safety mechanism.
1
It understands the intent. It reads your message and
decides what needs doing (e.g. "send email to Mr. Rossi").
2
It prepares the action. It writes the email in its
internal memory, but does not send it. If it's about downloading a
file, it computes the download but does not write it yet.
3
It asks for confirmation. It shows you
what it will do, where, with buttons
"Yes / No / Let's edit".
4
It acts only after "Yes". If you don't reply within
the expected time (2 minutes on Telegram, 30 seconds on the terminal),
the action expires: it does not happen. Nothing ever acts "in
the dark".
5
It confirms the outcome. It says "done" or "failed,
reason: X". Everything ends up in a write-only register: you can always
re-read yesterday, last week, six months ago.
Too many "may I?" getting annoying? Normal. For things
you do often, Mykleos learns: the first time it asks, the second time it
asks with "okay for similar things in the next 30 minutes?", and if you
say yes it stops asking for half an hour. If the channel changes (you're
at the computer instead of on the phone), it starts from scratch:
different contexts, different discretion.
5. Its memory: what it remembers and what it doesn't
Mykleos has a three-layer memory. It's not magic,
they're three files/tables with different roles.
Figure 3 — The memory. On the left, the warm and volatile; on the right, the cold and permanent.
The detail that matters: Mykleos never promotes a fact
to "long memory" on its own. Once a day (in the evening) it proposes:
"I noticed you talked often about Mr. Rossi today: shall I save him as
a regular contact?". You say yes or no. Long memory is built this
way, never behind your back.
6. How it grows: new tools when they're needed
Mykleos is not born "omniscient". It ships with a limited toolbox (read
files, send mail, download attachments, a handful of others) and learns
to widen it.
The idea is simple but powerful: when you ask it for something it
doesn't know how to do well (or at all), instead of giving up, it
proposes to build itself a new tool. A script, a small
wrapper, a reusable recipe. It shows you what it's about to create, you
give the OK, and from then on that piece joins its toolbox.
1
It tries with the tools it has. If a task fails, or is
painfully slow, it notices.
2
It proposes a new "actuator". "I notice that every time
you ask me to file invoices I do 6 manual steps. May I build a tool
file_invoice that does them in one?"
3
You approve (or not). If yes, Mykleos writes the new
tool, tests it in sandbox, adds it. If no, it forgets. The initiative is
its, the control is yours.
4
It keeps the ones that work. Tools you use often and
that succeed, stay. Those never used, or that fail, are proposed for
removal.
Biological analogy. It's like a little brain that grows new
synapses in front of recurring problems, and lets old ones die
when no longer useful. The document
Neurons, Synapses and Memory
tells how this really works, with examples. For now it's enough to know that
Mykleos doesn't stay frozen at day-one capabilities.
7. What it can do, what it will never do
✓ What it can do (or will be able to do soon)
Read and summarise your emails, draft replies
Search files on the home computer, extract data from PDFs and documents
Download attachments from cloud services (Drive, Dropbox)
Manage the calendar: add events, detect overlaps, suggest reschedules
Keep a shopping list / a to-do list
Answer your questions about yesterday, last week, months ago
Remember birthdays, anniversaries, deadlines
Read the web when useful (weather, train times, specific news)
Proactive briefings: send you spontaneous messages at set times (morning, Sunday, end of month)
Delegation to home automation: ask the home agent to check sensors, lights, appliances
Multi-family use: your partner / kids can have their own channel, with permissions different from yours
Answer you even from outside home, via secure tunnel (optional, switchable off)
Write and run small scripts in sandbox, for repetitive chores
Build itself new tools when its current ones aren't enough (always with your approval)
✗ What it will never do
Act on the internet without your explicit OK
Spend more than €5 in a day (hard cap)
Send messages to people not explicitly whitelisted
Delete files without showing them to you first
Leave the house (it does not control locks, alarms, cars)
Read your private data without your having given them to it
Modify its own "Laws" (they are Merkle-signed, sealed)
The 4 Laws. Mykleos operates under 4 rules it cannot
rewrite on its own: (0) stay within the agreed perimeter, (1) do no harm,
(2) obey only informed instructions, (3) always leave a trace of
everything. They are written in a digitally signed file: if anyone
tampers with it, Mykleos does not start. It is the safety belt of the
system.
8. How to get started (in practice)
Getting it up and running takes an afternoon. The pieces are:
The mini-computer. An Intel N100 at around
€150 is more than enough. Silent, 6-8 W.
Ubuntu + installer. A guided installer sets
everything up (the user's part is answering two or three
questions).
First Telegram pairing. You scan a code from your
phone, and the phone gets "paired" to your mini-computer. Everything
stays in the family: Telegram is the channel, not the brain.
The first three personalisation files. A file
with who you are ("my name is Roberto, I live in X, my alarm is at
7 a.m."), one with your contacts ("my lawyer is studio.rossi@…"),
and one with things to know ("work attachments go to
~/Documents/work/"). They take ten minutes to write, they can be
edited at any time.
Beginner's mode. You start in "supervised": it
asks confirmation for everything. After 10-15 days, once you
have built confidence, you unlock "autonomy full" on specific things
(e.g. "you may read emails without asking").
Estimated time to first "wow". One hour for the
installation, another one for the first useful scene (typically "summarise
my emails"). The interesting things, however, arrive after 2-3 weeks,
when the memory has started to know who you are.