The Deciduous Tutorial
A narrative guide to decision graph tooling for AI-assisted development.
Work smarter with careful rigor.
LLMs generate complex code fast. Reviewing it, understanding it, and maintaining it? That's on you. Deciduous gives you a structured record of every decision—what was considered, what was chosen, and why.
See It In Action
Watch how deciduous captures a decision in seconds:
Quick Demo: Capturing a Decision
user
deciduous add goal "Add user authentication" -c 90
output
Created node #1: goal "Add user authentication"
user
deciduous add decision "Choose auth strategy" -c 85
output
Created node #2: decision "Choose auth strategy"
user
deciduous add option "JWT tokens" -c 80
user
deciduous add option "Session cookies" -c 75
output
Created node #3: option "JWT tokens"
Created node #4: option "Session cookies"
user
deciduous link 1 2 -r "Need to decide auth approach"
user
deciduous link 2 3 -r "Stateless, good for API"
user
deciduous link 2 4 -r "Simpler but requires server state"
output
Created edge: #1 -> #2 (leads_to)
Created edge: #2 -> #3 (leads_to)
Created edge: #2 -> #4 (leads_to)
user
deciduous nodes
output
ID TYPE STATUS TITLE
#1 goal active Add user authentication
#2 decision active Choose auth strategy
#3 option active JWT tokens
#4 option active Session cookies
What You'll Learn
This tutorial will take you from zero to productive with deciduous. By the end, you'll know how to:
- Install and initialize deciduous in any project
- Create decision nodes and connect them into a graph
- Use deciduous with Claude Code for persistent AI memory
- Recover context when sessions end or compact
- Generate PR documentation from your decision graph
- Share decisions across a team with the patch system
- Use all three viewers: TUI, web, and CLI
The Workflow at a Glance
This is what working with Claude Code and deciduous looks like:
┌─────────────────────────────────────────┐
│ SESSION START │
│ Run /recover to restore context │
└──────────────────┬──────────────────────┘
│
┌──────────────────▼──────────────────────┐
│ USER REQUEST │
│ "Add rate limiting to the API" │
└──────────────────┬──────────────────────┘
│
┌───────────────────────────────┼───────────────────────────────┐
│ │ │
▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌───────────────┐
│ Log Goal │ │ Implement │ │ Log Outcome │
│ deciduous │ ──▶ │ Write code │ ──▶ │ deciduous │
│ add goal │ │ Run tests │ │ add outcome │
└───────────────┘ └───────────────┘ └───────────────┘
│ │ │
│ │ │
└───────────────────────────────┼───────────────────────────────┘
│
┌──────────────────▼──────────────────────┐
│ SYNC & COMMIT │
│ deciduous sync && git commit │
└──────────────────┬──────────────────────┘
│
┌──────────────────▼──────────────────────┐
│ SESSION END │
│ Context persists in graph │
└─────────────────────────────────────────┘
Who This Is For
This tutorial assumes you:
- Work with AI coding assistants (Claude Code, etc.)
- Are comfortable with the command line
- Have experienced the frustration of lost context
No prior knowledge of decision graphs or graph databases is needed.
How This Tutorial Works
The tutorial is structured progressively:
- Hello, World! gets you running in about five minutes
- Real-World Workflows shows realistic Claude Code sessions
- Viewing the Graph covers TUI, web, and CLI interfaces
- Sharing with Others explains multi-user sync
- Advanced Topics covers power features
- Plumbing Reference documents every command
You can read it straight through, or jump to the section you need.