Building
your Jarvis
It’s freezing outside. In here, you ignite your own intelligence — a multimodel assistant with memory and security. You don’t subscribe to it. You build it. And it warms you up.
Frameworks are bloated. Your Jarvis doesn’t have to be.
Most agents trap you in a tower of dependencies and a black box — cold and closed. Here, it’s the opposite: a feature store. You choose what goes in, understand each component, and Claude Code builds alongside you — live, in 2 days. The warmth comes from building.
Five steps until the spark catches
Select ingredients
Your Jarvis’s features
Generate document
ingredients.md
Switch to Claude Code
"Read and build"
Answer questions
Docker? Which model?
Assistant powered on
On Telegram, now
What you’ll build
A real assistant — warm, fast, and fortified. Not a demo chatbot.
In your channels
Telegram, WhatsApp, Discord, and voice.
Multiple models
OpenRouter (100+) and local Ollama. Switch the brain without rewriting.
Your own memory
SQLite FTS5 + BM25: it remembers you and the context.
Zero-trust security
Anti-prompt injection, crypto, sandbox, and gates.
Six tracks, from zero to expert
Independent, yet connected to the whole. From philosophy to advanced AI security.
Fundamentals
No-bloated-framework philosophy, the 5-component architecture, and setup.
Identity and Channels
SOUL.md, profiles, Telegram, WhatsApp, Discord, and SQLite memory.
Zero-Trust Security
Prompt injection, Fernet+PBKDF2, WASM sandbox, gates, and audit trail.
Memory and Integrations
OpenRouter, Ollama, Google Workspace, GitHub, and webhooks.
Advanced Architectures
6 Claw patterns, TinyClaw multi-agent systems, Agent Zero, and continuous learning.
Final Project: Your Jarvis
Planning, ingredients.md, Docker deployment, monitoring, and roadmap.
Eight feature tracks
Identity
SOUL.mdSecurity
WASM, gatesChannels
WhatsApp, voiceMemory
SQLite, BM25Automation
Cron, agentsIntegrations
OpenRouter, MCPSkills
Google, GitHubArchitectures
6 Claw patternsWho it's for
- → People who want an AI assistant your, not another subscription.
- → Developers and curious learners who want to understand agents for real.
- → People who already use AI and want to go from "using" to "building."
- → People who care about privacy—and want to run it locally.
Prerequisites
- · A computer (Windows, Mac, or Linux) and internet access.
- · Willingness to work in the terminal — we’ll guide you from the beginning.
- · You don’t need to know how to code. Seriously.
- · A Telegram account to test.
Choose how to warm up
2 days live, from zero to your Jarvis powered on. Limited spots.
- ✓ 2 days of live workshop
- ✓ Access to all 6 tracks and materials
- ✓ Guided assistant building
- ✓ INEMA.VIP Community
- ✓ Everything online, in person
- ✓ Hands-on guidance, right by your side
- ✓ Networking with the group
- ✓ Leave with the assistant up and running at the end
Registration and payment through INEMA.VIP
Frequently asked questions
Do I need to know how to code? +
No. We start from zero, including setup. Claude Code builds alongside you; your role is to decide what the assistant does.
What's the difference between online and in-person? +
The content is the same. In person, you get hands-on guidance by your side and networking. Online, you participate live from wherever you are.
Do I need to pay for AI models? +
You’ll learn to use OpenRouter (pay per use, just cents) and local models with Ollama (free, on your machine).
Will I leave the workshop with something ready? +
Yes. The goal is to finish with your own assistant working on Telegram, with the features you chose.
Where can I see the full content? +
The detailed syllabus is at inematds.github.io/intelecto-curso.
This winter,
light up your Jarvis
Stop subscribing to other people's intelligence. Build yours—and keep it warm.
ONLINE R$487 · IN-PERSON R$1987