Someone has to build it
Models, access, memory, webhooks, queues. Half the decisions are made before the agent ever answers once — and each of them later either holds the system up or breaks it.
>_ a turnkey assistant in telegram
Your assistant in Telegram. You write — it does the work.
Not a chatbot that answers nicely, but an agent that executes: it works through your inbox, keeps the calendar, files leads into Notion, places calls on your behalf and reports back in the same chat. One line from you is enough.
- ✓live in 1–2 days
- ✓runs 24/7
- ✓no code, no terminal
↗ You hand over the task in words or by voice. The agent closes the rest.
9:41
ai.architecttask, live
14 s
“invoice from the courier — put it in the register and remind me tomorrow if it goes unpaid”
Filed in the register. Lead in Notion. I'll remind you tomorrow at 10:00.
without you touching it200 OK · 4.1 s ·
- live in 1–2 days
- runs 24/7
- no code, no terminal
- live in 1–2 days
- runs 24/7
- no code, no terminal
- live in 1–2 days
- runs 24/7
- no code, no terminal
// the problem
Everyone wants an AI agent. It works for a handful.
Between “I want an agent” and “the agent clears my routine every day” stand three walls.
And it has to stay up
A demo works once. Production has to work daily: retries without duplicates, idempotency, failures visible immediately rather than a week later in silence. This is exactly where a toy stops being a tool.
And there is no time for it
You have a business and a queue of tasks. Becoming an AI engineer for two weeks is not a plan. You need a result that works tomorrow, and a person answerable for it.
All three walls are my job, not yours
I take on the architecture, memory, integrations, monitoring and support. You hand tasks to Telegram and get them done.
// the solution
One line from you. The rest is on me.
The assistant lives inside your own Telegram bot, on my infrastructure. You install nothing and keep nothing running.
what I need from you
Your Telegram bot token
1234567890:AA••••••••••••••••••••••••••••••••
You hand it over through a secure channel after we start — not in chat and not in a form. The bot stays yours: access can be revoked with one button.
from there it's me
- ✓Architecture, not a script. n8n + Claude Code, every link with error handling and its own branch for failure.
- ✓Real memory. It remembers you, your tasks and what was agreed between sessions — not just the last five messages.
- ✓Skills built around your processes. Assembled for what you actually do every day, not a universal assistant for everything.
- ✓My infrastructure. It lives with me, runs 24/7 and updates without your involvement.
- ✓Monitoring and support. I hear about an outage before you do — that is the deal, not a bonus.
// how it works
From a conversation to a working agent — three steps
We go through your tasks
We look at what eats your day and what can be handed over. You leave with a list of skills and a tier — no “we'll build everything and figure it out later”.
I build and test it
I set up memory, skills, access and integrations, then run it on your real data — demo examples break in places real life never does.
You delegate
You message the agent in Telegram like a colleague. Fixes and new skills are part of the subscription — no separate invoice for every small thing.
Average time from “we agreed” to a working agent — 1–2 business days.
// capabilities
What the assistant can do
It does more than answer messages: it uses tools, carries the task to the end and comes back with the result.
capability 01 / 10
Conversation
Write or send a voice message — in a private chat or a group. The agent holds the thread of the conversation, asks when a task is ambiguous, and answers like a person rather than a template.
text and voicegroup chatsvoice repliesEnglish · polski · українська · русский
Pick a capability on the left — the panel updates
// already running
On my own processes first, then with clients
Below are not promises but things that already run daily without a human.
of manual entry removed
Scanned invoices land in a Drive folder. The agent reads each one, pulls out the details and fills the register in Google Sheets — with no typing.
to the first reply to a lead
A form submission notifies you instantly and sends the client a personal reply in their own language. No request waits until morning.
answers without a human
An online shop's email agent reads messages, checks stock and prices, routes orders and replies — like support that never sleeps.
// my own tool, not a demo
The assistant I use every day
I throw a task at it by voice — n8n turns it into a calendar event, a note in Notion, a task in Telegram or a reminder that fires days later, and reports out loud what exactly happened. This is my working tool, not a page for a website.
voice→transcription→intent and slots→Notion · Calendar · Sheets→Telegram→spoken reply
// pricing
Choose how much work to delegate
Development is one-off; the subscription covers the agent's work, infrastructure and support. You can move up a tier at any time.
Start
An assistant for everyday routine
Development$22Subscription$99/mo
Choose Start- ✓Text and voice in Telegram, private chats and groups
- ✓Long, short and contextual memory
- ✓Gmail, Calendar, Docs, Sheets, Drive
- ✓Reads files and documents
- ✓Web search
- ✓Tasks, reminders, daily digest
- ✓Monitoring and support
Optimal
Content, research and deeper automation
Development$99Subscription$149/mo
Choose Optimal+ everything in Starton top- ✓Answers by voice
- ✓Generates images and cards for social media
- ✓Turns YouTube, Reels and TikTok into summaries
- ✓Deep web research with a source list
- ✓Keeps your spreadsheets and tracks goal progress
- ✓Personal analytics and reports
Max
Communication, sales, integrations
Development$199Subscription$199/mo
Choose Max+ everything in Optimalon top- ✓Places calls on your behalf and takes incoming ones
- ✓Instagram and Threads: collects requests from DMs
- ✓Connects to your CRM
- ✓A transcript of every conversation in your register
- ✓Priority support and changes
Custom
from $1000An agent for a specific business process: your integrations, your rules, several roles and your own data perimeter.
// comparison
ChatGPT answers. An agent does.
A chat helps you find an answer. An agent carries the matter to a result and reports when it is closed.
| Criterion | ChatGPT Plus$20/mo | Build it yourself$100+/mo and your time | Your agentturnkey |
|---|---|---|---|
| Ready to work immediately | partly | no | yes |
| Executes tasks instead of advising | no | you build it | yes |
| Remembers your context and agreements | limited | you configure it | yes |
| Lives in your Telegram | no | you configure it | yes |
| Integrations with mail, calendar, CRM | limited | you build it | already connected |
| Technical skills required | no | yes | no |
| Who keeps it running 24/7 | you | you | me |
| Support and changes | no | no | included |
// request
Ready to delegate this week?
Leave your Telegram — we'll pick a tier, assemble the skills around your tasks and launch the agent. No code, no terminal, no “figure it out yourself first”.
↗ Or just write to @ViiacheslavM
// faq
Frequent questions
Is this safe?Security
The agent runs inside your own bot and only touches what you gave it access to. You hand over the token through a secure channel after we start, not in a form on a website. Your data is not used to train models and is not passed to third parties outside the list of services you see before we begin.
What do you see from my conversations?Security
Technical execution logs — that is what I debug failures with. Access to message content can be restricted at the start: diagnosis then takes longer, but privacy is complete. This is decided once and put in writing.
Where is my data stored?Security
On my infrastructure and in the services the agent uses to work — mail, calendar, the notes database. The full list of services, regions and retention periods comes before payment, not after.
Can I delete everything?Control
Yes. Memory and data are deleted on request, you revoke access yourself in your accounts, and the bot stays yours. Whatever the agent accumulated I hand back in a readable export.
Will the agent do something without my permission?Control
Anything visible from outside — an email to a client, a call, a publication — goes through your confirmation in chat by default. Only what you explicitly allow becomes autonomous.
How long does launch take?Launch
Usually 1–2 business days after the conversation and access. The Max tier with calls and CRM takes longer, because it needs phone numbers and setup on your services' side.
What do you need from me at the start?Launch
Half an hour of conversation, a Telegram bot token and access to the services the agent has to work with. Everything else is my part.
What happens if I stop paying?Payment
The agent is switched off, I hand the data back as an export, and the bot stays with you. No hidden holds and no forced “last month”.
Can I move the agent to my own setup?Payment
Yes. I hand over the build, you deploy it on your own infrastructure and keep it running yourself. Support in that case is scoped separately — write to me.
What will the agent not do?Limits
It will not replace a lawyer, an accountant or a doctor — it prepares the material, the decision stays with a human. It does not work with services that have no API. And it does not give hundred-percent accuracy: where the cost of a mistake is high, your confirmation stands in the way.