Build an AI Second Brain That Remembers Why You Decided

"Why did we decide that?"

You remember what you chose. CSV export instead of a live dashboard. Postgres instead of MongoDB. Weekly releases instead of daily. But why? That answer is buried in a conversation from three months ago.

This guide shows you how to build a second brain for your decisions with your AI agent and Ragionex Memory, using a template you can copy right now.

The missing piece is the reason

"Use CSV for customer reports" tells you what you chose. It does not tell you whether that was a deadline compromise, a customer request or a permanent decision.

When the situation changes, the outcome alone is useless. You need to know which constraint led to the choice, so you can tell whether that constraint still holds.

So save the decision, the reason and the condition for reconsidering it together, in one memory.

The Decision Memory Template

Copy this, fill it in while the reasoning is fresh, and send it to your agent:

Save this decision in Ragionex Memory. Keep the reason and the
revisit condition together. Ask me about missing details
instead of inventing them.

Context:    [What this decision applies to.]
Decision:   [What I chose.]
Reason:     [Why, including the key constraint.]
Trade-off:  [What I knowingly gave up.]
Revisit if: [What would make me reconsider.]

Each line answers a question your future self will ask:

The Reason line matters most. Write the real constraint. If it was a deadline, name the deadline. If you do not remember the reason, write "unknown". An honest gap beats an invented reason.

A filled-in example

A made-up decision from a made-up project:

Save this decision in Ragionex Memory:

Context:    Customer reports for the Acme pilot.
Decision:   Offer a downloadable CSV report before building a live dashboard.
Reason:     Pilot customers review figures weekly in a spreadsheet.
            Nobody asked for live data. CSV fits that workflow
            without building a reporting screen.
Trade-off:  Customers download and open the file themselves.
Revisit if: Customers need current figures during the day.

Months later: ask "why?"

Open a fresh session. Do not paste the decision back. Ask:

Check Ragionex Memory: why did we choose CSV over a live dashboard
for the Acme pilot? Show me the saved decision.

In this example, the agent finds the saved decision and answers from it:

CSV matched the pilot customers' weekly spreadsheet review. Nobody asked for live data, so the reporting screen was deferred. Reconsider if customers need current figures during the day.

You get the reason, the trade-off and the trigger for change, straight from what you saved. Ask for the saved text and you can check every word.

When the situation changes

This is where a decision memory pays off. Suppose customers now want live figures:

Customers now need current figures during the day.
Recall the Acme reporting decision from Ragionex Memory.
Which original constraint has changed?

The agent compares the new requirement with your saved decision and finds the match: the "Revisit if" condition has been met. Your next discussion starts from the original reasoning, not from a blank page.

And when you make the new decision, update the memory so the reason stays current. See the memory repair guide.

Set it up

You need Node.js 18 or later and an AI agent, such as Claude Code. Run the setup:

npx.cmd -y @ragionex/memory-mcp@latest install

npx -y @ragionex/memory-mcp@latest install

Pick your agent if asked, then quit and reopen it. Step-by-step details: the Claude Code first-install guide. Using several agents? They can all share one second brain: connect them to one shared memory.

Start free

The Free plan includes 10,000 requests every 30 days and memories up to 10,000 characters, enough room for a full decision with its reasoning. Browsing and reading your memories in the dashboard is not counted as a request.

Set it up in one line, copy the template above, and save the first decision you know you will be asked about again. Full reference: the Memory quickstart.