Mined from real sessions · not written by hand

Your AI agents act like they just met you. Emulo fixes that.

It mines your real Claude Code, Codex, Cursor and Copilot sessions into a you.md your agent reads before every task. Not what you told it. What your work already proved.

★ Star on GitHub $ npx skills add ohad6k/emulo@emulo

or pip install emulo to run it yourself

Emulo engraving: a human and a machine head back to back
0
Sessions mined
0.0M
Tokens of just me
0
Months of history
I.

Not memory.

Memory

What you told the model

Your CLAUDE.md, your rules file, the preferences you remembered to write down on a good day.

  • Written once, goes stale
  • Only the rules you noticed
  • How you think you work
Emulo

What your work proved

Mined from the raw session logs: every rejection, every "that's not done", every time you asked for proof.

  • What you reject on sight
  • What "done" actually means to you
  • How you talk when you're really working
II.

The rules nobody wrote down.

Laws · ranked by how many of the 20 mining agents found each 17/20 consensus
LEX I Done means it runs live. Never trust "done" off a code edit. 17/20
LEX II Fix the one thing. Rewriting what isn't broken gets rejected. 15/20
LEX III No walls of text. Short, concrete, no hype. 14/20
The uncomfortable one asks for proof more the closer a task gets to shipping, and never on day one

This is one person's mine. Yours is pulled from your own logs and will read nothing like it.

III.

How the mine works.

1

Reads your history

Pulls every message you typed from your local Claude Code, Codex, Cursor, Copilot CLI and OpenCode logs. Strips tool output, pasted errors, file dumps. Keeps only your words.

2

Mines the patterns

Agents read each slice of your history and pull the repeated traits: how you define done, what you reject, when you ask for proof, how you actually talk.

3

Writes you.md

The repeated traits merge into one profile, with dated receipts from real sessions. Installs as a skill, so your agent reads it before every task.

Emulo reads your private session logs, so it is built to never move them. Everything runs on your machine, secrets are redacted before a single line is written, and the profile stays in a local file you can open, edit, or delete.

IV.
Free forever, and one thing I do by hand

The tool is free. The judgment is the part you pay for.

Emulo runs on your machine and stays free and MIT forever. What it cannot do is tell you which of the rules it mined are actually true about you. That pass is judgment, and I do it by hand.

Emulo

Free and open source.

Mine your own sessions and keep the resulting you.md on your machine. The core product stays useful without a cloud account.

$0forever
  • Local session mining
  • Editable, portable you.md
  • MIT licensed source
  • No subscription
Get Emulo
Profile Build

I read your receipts and cut what cannot be defended.

Mining takes seconds. Deciding which of the mined rules are actually true about you, and cutting the rest, is the part that takes judgment, and it is the part most people never finish. I do that pass by hand.

  • Every rule whose quote appears in none of your sessions removed, not softened
  • Every rule that would be true of any developer removed as well
  • What survives restructured into work, design, write and video layers
  • The exact install lines for the agents you actually use
  • A written read of what your own logs say about how you work the model
Profile BuildBack in three business days
$300 one time

You send two files, both written by the free tool on your own machine: python emulo.py for the draft, then python emulo.py --coach --json for the receipts. I never ask for your session logs and I never see them. If it does not contain at least three rules you agree with and would never have thought to write down, I refund you and tell you why. Paid by invoice or PayPal, so it does not go through a store. ohadkrispin@gmail.com

Claude Code· Codex· Cursor· Copilot CLI· OpenCode·
IV.

Questions.

How do I make my AI agent write code the way I do?

Give it evidence instead of instructions. Your session logs already record how you decide, what you reject, and what you call done. Emulo reads those logs, mines the repeated patterns, and writes a you.md your agent loads before every task, so it starts from your standards instead of a blank slate.

How is this different from agent memory?

Memory stores what you told the model. Emulo mines what your work already proved about you, which is a different input entirely.

That is why it reads raw session logs and not your CLAUDE.md. The rules you remembered to write down on a good day are a small and flattering subset of how you actually work.

Is this just a CLAUDE.md or a rules file?

No. A rules file is written by hand, so it captures intentions. Emulo's profile is derived from real sessions, with dated receipts pointing back at the ones that support each trait, so you can check any claim it makes about you.

Do my session logs leave my machine?

Session extraction, redaction, caching, and the profile itself all stay local, and secrets are redacted before anything is written. The one exception is mining: if you point it at a hosted model, the selected evidence goes to that provider. Choosing a local model keeps the whole run on your machine.

Which coding agents does it work with?

It reads Claude Code, Codex, Copilot CLI, OpenCode, and Google Antigravity logs out of the box, and installs profiles for Cursor and Gemini through explicit adapters. It also ships an MCP server, so any MCP client can load the profile before a task.

What does it cost to run?

The tool is MIT licensed and free. Mining spends model calls to read your history, so Emulo prints a read-only plan with the cost first and waits for your approval before any model work starts.

How much history do I need?

It mines whatever you have. Patterns get sharper with more sessions, since a trait only earns a place in the profile when separate sessions keep supporting it. The profile on this page came from 1,656 sessions over nine months.

Image

Stop starting from zero.

One command. Your logs become the profile your agent reads first.

★ Star on GitHub $ npx skills add ohad6k/emulo@emulo

or pip install emulo to run it yourself