Copana — Personal AI, powered by markdown
Open source · MIT License
Copana - OpenClaw alternative that works with Claude Code | Product Hunt

Copana:
Personal AI, powered by markdown.

Copana is a personal AI assistant that actually knows you. Memory, personality, and context — all in plain text files you can read, edit, and own.

terminal 
$ git clone https://github.com/PayRequest/copana.git
$ cd copana && ./setup.sh
$ claude
# Your AI remembers everything. Forever.
Get Started See how it works

Every other AI tool

  • ✗ Forgets you the moment you close the tab
  • ✗ No context about your goals, preferences, or history
  • ✗ No follow-through on commitments
  • ✗ Your data lives on someone else's server

Copana

  • ✓ Remembers everything across sessions
  • ✓ Learns your preferences silently over time
  • ✓ Follows up on open loops and commitments
  • ✓ Plain markdown files you own in a git repo

How it works

Copana is a template repo for Claude Code. Your AI reads markdown files for context, and updates them as it learns about you.

1

Clone and setup

Run setup.sh to personalize your AI. Pick a name, a vibe, and you're done.

$ git clone https://github.com/PayRequest/copana.git my-ai
$ cd my-ai && ./setup.sh

Your name: Alex
Your AI's name: Atlas
Pick a vibe (1-4): 1

  Setting up Atlas for Alex...
  ✓ CLAUDE.md
  ✓ soul.md
  ✓ personality.md
  ✓ memory.md
  Setup complete!
2

Start talking

Run claude and your AI picks up right where you left off. Every session starts with a startup ritual — reading memory, checking tasks, getting the time.

$ claude

(atlas) [Monday 03 March, 09:15]

Morning. You've got 3 pending tasks and an open loop
about the dentist appointment from last week.

Your top priority is finishing the API docs — you
said Friday you'd have them done by Tuesday.

Want to start there?
3

It learns about you

Your AI silently captures facts as you talk. After a few sessions, it knows your work style, preferences, goals, and patterns.

user.md (after 5 sessions)
- Work: Software engineer
- Stack: Python, FastAPI
- Enjoys: Running, cooking
- Hates: Meetings
- Goal: Launch side project by Q2
preferences.md
- Morning: Coffee first, then work
- Communication: Slack > email
- Work style: Deep focus blocks
- Hates: Context switching
- Planning: Action > planning

Everything you need

No infrastructure. No Docker. No build step. Just markdown files and Claude Code.

🧠

Persistent Memory

Core facts, open loops, session logs, and decisions — all in structured markdown that persists forever.

✨

Personality

Give your AI a soul. Choose its vibe: casual buddy, professional partner, or tough coach. Defined in soul.md + output style.

🔍

Active Learning

Your AI silently captures preferences, habits, and opinions. No prompting needed — it just listens.

🤖

Subagents

Specialized AI agents for memory cleanup, web research, and daily reviews. Restricted tools, persistent memory.

📊

Status Line

Always-visible status bar showing open tasks, loops, and last session. Context at a glance while you work.

🔔

Proactive Nudges

Smart reminders about stale tasks, birthdays, and open loops. Optional cron-powered automation scripts.

🔒

Privacy First

Everything stays on your machine. Plain text files, version-controlled with git. No cloud, no telemetry.

📱

Multi-Device

Mac via Claude Code, mobile via Telegram bot or Claude app. Git keeps everything in sync.

🧩

Skills System

Add capabilities with slash commands: /add-telegram, /add-fitness, /add-budget. Create your own with /add-agent.

Just markdown files

No database. No embeddings. No vector store. Your AI's entire context is plain text you can read, edit, and version-control.

copana/
├── CLAUDE.md              # AI instructions (auto-read)
├── soul.md                # Identity & values
├── personality.md         # How the AI talks
├── user.md                # About you (AI fills this in)
├── memory.md              # Long-term memory & session log
├── tasks.md               # Active tasks
├── insights.md            # Strategy & follow-ups
├── routines.md            # Daily patterns
├── contacts.md            # Important people
├── .claude/
│   ├── settings.json      # Status line config
│   ├── agents/            # Subagents
│   │   ├── memory-manager.md
│   │   ├── research.md
│   │   └── daily-review.md
│   ├── output-styles/     # Personality style
│   └── skills/            # /setup, /add-telegram, etc.
├── scripts/               # Optional automations
└── docs/                  # Full documentation

Built for Claude Code

Copana uses Claude Code's most advanced features — subagents, output styles, and status lines. Not just a CLAUDE.md file.

.claude/agents/

Subagents

01
memory-manager — Consolidates files, archives old sessions, deduplicates facts
02
research — Web research with structured output, sources cited
03
daily-review — End-of-day review, pattern detection, daily journal

Restricted tools · Persistent memory · Create your own with /add-agent

.claude/settings.json

Status Line

Tasks: 5 | Loops: 3 | Session Feb 26

Always-visible context at the bottom of your terminal. See open tasks, unresolved loops, and when your last session was — without asking.

Auto-refreshes every 30 seconds

.claude/output-styles/

Output Style

Your AI's personality as a proper Claude Code output style. Direct, opinionated, no corporate fluff.

# Things you never say:
"Great question!"
"I'd be happy to help!"
"Certainly!"

Clean separation from CLAUDE.md instructions

Copana vs NanoClaw vs OpenClaw

Copana NanoClaw OpenClaw
Source files 50 15 3,680
Lines of code ~3,600 ~3,900 434,453
Dependencies 5 <10 70
Config files 1 0 53
Time to understand 8 minutes 8 minutes 1–2 weeks
Security model OS container isolation OS container isolation Application-level checks
Architecture Single process + isolated containers Single process + isolated containers Single process, shared memory
Setup git clone + setup.sh Docker + Node.js npm install + onboard wizard
Infrastructure None Docker required Node ≥22
Memory Structured markdown Session logs Per-agent sessions
Personality soul.md + personality.md None None
Personal context user.md + preferences.md None Workspace config
Active learning Captures facts silently None None
Subagents 3 built-in + custom None Multi-agent routing
Channels CLI + mobile (git sync) CLI WhatsApp, Telegram, Slack, Discord, Signal, iMessage + 6 more
Extensibility Skills + agents Skills Skills + tools + channels
Privacy Local files, git-versioned Container Local gateway (127.0.0.1)
Token footprint ~14k tokens (7%) ~35k tokens (17%) —

Get started in 30 seconds

Clone the repo. Run setup. Start talking. Your AI gets smarter with every conversation.

$ git clone https://github.com/PayRequest/copana.git
$ cd copana && ./setup.sh
$ claude
View on GitHub Star on GitHub