Developers have git. Now the rest of us get a history too.
Work done in Claude leaves a document and no trail. Let your agent record each decision as you go, and anyone can later see why things went the way they did.
You spend an hour in Claude writing a refund policy. On the way you make five decisions.
Tomorrow the policy exists. The decisions are gone. They sit in a chat you will never scroll back through.
Three months later someone asks: "Why did we do it this way?" You try to remember.
Developers fixed this years ago
They use a tool called git. It saves every change to the code, with who made it, when, and a note on why.
A developer can go back a year and see why a line was written. A new hire reads the history and catches up.
The rest of us never had that. We have the final document, and "ask Elena, she might remember".
Now your agent can keep the history
You do not write it. Your agent does, while it works with you.
It writes at four moments:
- Start. What it is about to do, and why.
- Decision or pivot. What it chose and why, or how the plan changed. Also what it ruled out.
- Blocker. What stopped it.
- Done. What it made, where it is, and what is still open.
You set this up once. After that it just happens.
What it looks like
Elena is drafting a refund policy in Claude. Her agent logs as they go:
She did not type any of that. She was working on the policy.
Three months later
A customer asks a teammate why there is no 14-day cutoff. She asks, and gets the reason with a link to the trail:
No meeting. Nobody had to remember.
What you get
- "Why did we do this?" has an answer in a minute.
- Rejected ideas stay rejected. Nobody proposes them again. Neither does the next agent.
- New people catch up on their own. They read the history.
- You can go on holiday. The reasons are no longer only in your head.
One difference from git
Git records what changed. This records why.
What to use
- A "decisions" doc. Ask your agent to add a line after each decision. Free, and fine for one person. You have to remember to ask, and it only covers that one doc.
- A decision log in your wiki. It works as long as people remember to update it.
- Naumu. This is what we build. Each piece of work becomes a thread in a topic your team follows. People read it. Agents search it. Anyone can ask why.
Set it up with Naumu
- Create an account and a space.
- Open Settings → Your agents and pick your app.
- For Claude or ChatGPT, add the URL it shows as a custom connector. Then sign in in the browser. There is no API key.
- Pick the topic the log should go to.
- Copy the setup prompt and paste it into a new chat. Accept when it offers to save the skill.
- Work as usual. Say "don't log this" to skip one.
Do you work in code? The same log works for coding agents.