I wanted OpenClaw's philosophy without the autonomy. So I built my own.

In the last two posts I explained why we built Second Brain and announced the launch. A lot of people replied asking about the technical side. How does data actually get in? What formats does it support? How does it connect to Claude? I’ll try to cover those questions here.

But first I want to talk about something that influenced how we designed this.

Why not just use OpenClaw

If you follow the AI space, you’ve probably heard about OpenClaw. It can send emails, browse the web, negotiate deals, create social media profiles, all while you sleep. People are building teams of 9 agents that run their work and personal life simultaneously.

I genuinely liked the idea. Local-first, open-source, persistent memory. The philosophy is right.

But I wasn’t ready to hand over my business data to an autonomous agent that runs 24/7 and acts on its own. Not because the technology is bad. Because the trust model isn’t there yet for me. My database has customer payment records, contact details, internal notes, financial data across three companies. The moment an agent starts acting autonomously on that data, sending emails I didn’t review, making decisions I didn’t approve, modifying records while I sleep, I lose control.

Many people are raising concerns about OpenClaw’s broad access to files and terminal. One user’s agent created a dating profile without his knowledge. These are early-stage problems that will get solved. But right now, for real business data, I wanted something different.

So we built Second Brain with a different philosophy: give AI full read and write access to your data, but only when you’re in the conversation. Nothing happens without you asking for it. Nothing runs in the background by AI, but you still have workflows running in the background. You stay in control of every query and every change, and you keep the track.

Same local-first approach. Same privacy model. But you’re the operator, not the passenger.

The database layer

Second Brain runs on a local database on your computer. No server, no cloud. You can open it in any database viewer, back it up by copying a file, or move it to another machine by dragging it over.

The schema is designed around how businesses actually organize information. Contacts, email addresses, phone numbers, organizations, relationships between people and companies, customer records with spend tracking, meeting notes, hiring and interview data, articles, documents, and structured notes. It’s not a generic “put anything in a table” setup. The schema has real structure so AI can reason over it properly. When Claude sees a contact with linked email addresses, payment history, organization affiliation, and meeting notes, it can answer questions that cross all of those dimensions.

The most important part is the application that runs locally. You can search, filter, browse, and manage your data without touching the command line.

Getting data in

This is the part that matters most, because a database is only useful if it’s easy to fill.

Second Brain has import parsers for multiple sources and formats. CSV files, ZIP archives, platform exports.

What you can import right now: contacts from any CRM or platform that exports CSV. Stripe transaction and subscription data. Customer records with spend data. Platform-specific exports like calendar events and more. Documents and files with text extraction.

The import pipeline doesn’t care where the data came from. If you can get it into a CSV or a supported export format, it goes into the database cleanly. We also implemented database optimization and cleanup functions.

Hexomatic integration

This is where Second Brain connects directly to what we’ve been building at Hexact for years.

Hexomatic was always a tool for collecting structured data from the web. Competitor pricing pages, Google Maps listings, job postings, product catalogs, lead information, news and articles.

Second Brain now connects directly to the Hexomatic API. You set up your scraping workflows in Hexomatic, define the schedule, and the results flow into your Second Brain automatically. No manual downloads. No copy-pasting between tools.

Your competitor changes their pricing? It’s in your database before your morning coffee. A new business opens in your market? Your local database gets updated. A job posting appears that signals a competitor is building a new team? Logged and queryable.

The combination of the two products creates something neither could do alone. Hexomatic collects. Second Brain stores, organizes, and makes it queryable through Claude. The scraping engine becomes the supply chain for your AI knowledge base.

The Claude connection

Second Brain connects to Claude Desktop through MCP (Model Context Protocol). This is the open standard that lets Claude interact with external data sources directly inside a conversation.

Once connected, Claude can read from your database and write to it. Ask a question in plain language, get an answer from your real records. Tell Claude to save a meeting note, update a contact, log a decision, it writes directly to the database.

The key design choice: Claude only accesses your data when you’re talking to it. You ask, it answers. You instruct, it writes.

You get the full power of a frontier AI model working with your complete business data, but you stay in the loop for every interaction. For business data, I believe that’s the right tradeoff. Maybe not forever. But right now, absolutely.

What this means practically

I just started using the chat as the universal interface to my CRM, spreadsheets, folders, and everything else.

A week in, you stop pasting context into prompts. A month in, you stop opening half the tools you used to switch between. The database becomes your operating layer. Claude becomes the interface. Claude by itself is already a great tool, it was just missing the context. Now it’s not.

That’s the system. Local, private, controlled, and connected to real data through Hexomatic pipelines that keep it current.

Second Brain launches April 20. Secure your license now at secondbrain.hexact.io.

If you already have a Hexomatic account, you get a free bonus session showing how to connect your scraping workflows directly to your Second Brain.