PTENES
Step-by-step guide

From Zero to Your Own
working assistant

You don't need to know how to code. Choose what you want, give the document to Claude Code, and it builds everything.

How INTELECTO was actually built

The 5 real steps

This isn't theory. It's exactly what happened to get INTELECTO from zero to working on Telegram.

1
Passed the reference document
The file my-framework-ingredients.md — generated by selecting features on the comparison page—was given to Claude Code as a starting point. This document describes everything you want the assistant to do.
2
Asked to generate a plan
One simple instruction: "read the document and give me a plan for how to build this." Claude read everything and returned a structure of workstreams, files, dependencies, and execution order — before writing a single line of code.
3
Claude asked questions during planning
Docker or direct installation? OpenRouter or Ollama only? Which default model? Claude identified the points that depended on your decision and asked before assuming. You answered in Portuguese. It adjusted the plan.
4
Executed the plan
With the plan approved, Claude created each file in the right order: config, security, providers, memory, Telegram channel, agent loop, setup wizard, and launcher. It showed what it was doing in real time and asked for confirmation before important actions.
5
You copied the .env file and started Docker
One action: copy the file .env with your API keys (Telegram and OpenRouter) and run it docker compose up. The assistant started up, connected to Telegram, and began responding.
The complete process
🛒
You do the shopping Check out the framework comparison and choose what you want: channel, memory, AI provider, security...
📄
The document is generated Your choices become a detailed specification file — the "blueprint" for your assistant.
💬
You give Claude Code an instruction One sentence. That's it. "Read the file and build everything."
⚙️
Claude Code builds It reads the document and writes each code file in the right order, asking for confirmation before important actions.
✅
You test and adjust Send a message on Telegram. The assistant responds. Ask for changes in Portuguese.
Detailed step-by-step guide
1
Do your shopping

Go to the framework comparison page and choose the ingredients that make sense for your day-to-day needs. No technical knowledge required — think in terms of features.

Examples of choices
  • ✓  Communication channel → Telegram
  • ✓  Memory → Stores everything I say (SQLite)
  • ✓  Brain (AI) → OpenRouter — access to 100+ models
  • ✓  Run on → Docker (any machine)
  • ✓  Personality → Editable Text Files
  • ✗  I don't need it → browser automation, task scheduling
Go to the purchases →
2
Receive the specification document

Based on your choices, a file is generated—let's call it meu-assistente.md — which describes in detail what needs to be built.

What this file contains
  • 📁  Which files to create and what each one does
  • 🔗  How the parts connect to each other
  • 📋  The order in which everything should be built
  • ⚙️  The specific settings for your choices
  • 🔒  The security rules Claude must follow
💡

You don't write this file

The document is generated automatically from your choices. You just read, review, and approve it.

3
Open Claude Code

Place the generated document in the project folder and open the terminal in that folder.

# Navigate to the project folder
cd minha-pasta-do-projeto

# Open Claude Code
claude
📦

Do I need to install Claude Code?

Just once: npm install -g @anthropic-ai/claude-code in the terminal. Then just use the command claude.

4
Give the instruction—one sentence

Inside Claude Code, you type a single sentence. The document already has all the details—you just tell Claude to read and build it.

The instruction you type

"Read the file my-assistant.md and build everything described in it"

That's it. Claude Code reads the document, understands the plan, and gets to work.

What happens after you send the instruction
  • 1. Claude reads the entire document
  • 2. Creates the files in the right order
  • 3. Shows you what it's doing in real time
  • 4. Asks for your confirmation before important actions
  • 5. Lets you know when it's done
5
Turn on the assistant

When Claude Code finishes, you start the assistant with a command.

If you chose Docker
docker compose up
If you chose direct installation (Mac)
./start.sh

Open Telegram and send a message to your bot. It replies.

6
Adjust anything you want

If something isn’t the way you wanted, just tell Claude Code in Portuguese.

Examples of adjustment requests

"I want it to always respond in Portuguese"

"Change the personality — I want it to be more direct"

"I want it to remember the names of people who messaged me"

There is no limit to adjustments. It is an ongoing refinement process.

Frequently asked questions
Do I need to know how to code?

No. You choose what you want in plain language, and Claude Code writes the code. You test it and ask for adjustments in Portuguese.

What is CLAUDE.md? Do I need to create one?

CLAUDE.md is an optional file that stores project rules so Claude doesn't need to be reminded every time. It is generated automatically along with your specification document. If it doesn't exist, you just need to be a little more specific in your instructions.

What’s the difference between the purchasing document and what Claude builds?

The shopping document is the plan — describes what you want. What Claude builds are the real code files — the assistant actually working.

What if Claude gets something wrong?

You describe the problem in Portuguese and ask for a fix. Claude rereads the code and fixes it. This correction cycle is normal and expected.

Docker vs. direct installation—which should you choose?
Docker Direct installation
Runs on any machine ✓ Yes Mac Only
Simpler to install Requires Docker to be installed ✓ Yes
Server / VPS ✓ Ideal Hard
Personal use on Mac Works ✓ Simpler

Ready to get started?

Compare the frameworks, choose your ingredients, and come back here to build.

Do my shopping → Back to the beginning