Copana's memory is structured markdown files. No database, no embeddings, no vector store. Just text files that both you and your AI can read and edit.
- Human-readable. Open any file and understand it instantly.
- Version-controlled. Git tracks every change. Roll back anytime.
- Portable. No lock-in. Move to a different AI tool if you want.
- Editable. Fix mistakes, add context, reorganize.
- Simple. No infrastructure to maintain.
The central memory file. Contains:
- Core Facts — Permanent knowledge (your name, job, family, etc.)
- Open Loops — Things to follow up on
- Recent Decisions — Important choices for consistency
- Session Log — What happened in recent sessions
Your AI reads this every session to maintain continuity.
Everything the AI learns about you. Filled in over time through conversation:
- Background and skills
- Personality traits
- Goals and motivations
- Work style
One-liners about your preferences:
- Communication style
- Tool preferences
- Work habits
- Life philosophy
Simple task tracking:
- [ ] Write blog post about X #work
- [ ] Call dentist #personal due:2026-03-01
- [x] Set up Copana (2026-02-26)Advice given, lessons learned, ideas backlog, and follow-ups. This is your AI's strategic memory.
Your daily/weekly routines, observed over time. Helps the AI know when to nudge, when to leave you alone.
Important people with context. Your AI references this when someone comes up in conversation.
Daily notes and reflections. Can be written by you or by the AI's daily reflection script.
user.md:
- Name: Alex
- Timezone: America/New_York
user.md:
- Name: Alex
- Timezone: America/New_York
- Work: Software engineer at a startup
- Enjoys: Running, cooking
- Hates: Meetings
user.md:
- Name: Alex
- Timezone: America/New_York
- Work: Software engineer at CloudCo (since 2024)
- Stack: Python, FastAPI, PostgreSQL
- Side project: Building a habit tracker app
- Enjoys: Running (training for half marathon), cooking (Italian food)
- Hates: Meetings, long Slack threads
- Morning routine: Coffee, run, then work
- Goal: Launch side project by Q2
The AI handles most maintenance automatically:
- Adds new facts as it learns them
- Updates outdated information
- Moves completed tasks
- Clears resolved open loops
You can also:
- Edit any file directly
- Delete incorrect information
- Reorganize sections
- Archive old session logs to
memory/archive/
- Keep
memory.mdconcise — it's read every session - Archive session logs older than 2 weeks
- Review
user.mdoccasionally — is it accurate? - Use
insights.mdfor things you want the AI to remember long-term