Deciduous
Decision graph tooling for Claude Code. Track every goal, decision, and outcome. Survive context loss. Generate PR writeups automatically.
How It Works With Claude
Deciduous gives Claude a persistent memory that survives context loss. As Claude works, it logs decisions to a graph. When a session ends or context compacts, the graph remains - queryable by the next session.
The Workflow
- Session starts → Claude runs
/contextto see past decisions - Claude works → Logs goals, decisions, actions as it goes
- Session ends → Graph survives for next time
- PR time → Generate writeup + visualization from the graph
What Claude Logs
Every step is captured in real-time:
Goal: "Compile rwc to WASM"
└─> Decision: "Use wasm-bindgen + wasm-pack"
└─> Action: "Add wasm-bindgen config"
└─> Action: "Add lib.rs exports"
└─> Observation: "Homebrew lacks wasm32 target"
└─> Decision: "Install rustup"
└─> Outcome: "WASM build successful"
PR Generation
Generate a full PR writeup with decision graph visualization:
&&
The writeup includes goals, decisions, actions, outcomes, and an embedded graph:


Interactive Web Viewer
View the full decision graph with deciduous serve:


Quick Start
1. Install
2. Initialize in your project
This creates:
.deciduous/- SQLite database.claude/commands/decision.md-/decisionslash command.claude/commands/context.md-/contextslash command.github/workflows/cleanup-decision-graphs.yml- Auto-cleanup PNGs after PR mergeCLAUDE.md- Project instructions with workflow
3. Start tracking
# Claude logs decisions as it works
# View the graph
Commands
# Core workflow
# PR workflow (recommended: use --auto for branch-specific filenames)
# Export
Node Types
| Type | Purpose | Example |
|---|---|---|
goal |
High-level objectives | "Add user authentication" |
decision |
Choice points | "Choose auth method" |
option |
Approaches considered | "Use JWT tokens" |
action |
What was implemented | "Added JWT middleware" |
outcome |
What happened | "JWT auth working" |
observation |
Technical insights | "Existing code uses sessions" |
Confidence Weights
Every node can have a confidence score (0-100):
| Range | Meaning |
|---|---|
| 90-100 | Certain, proven |
| 70-89 | High confidence |
| 50-69 | Moderate, some unknowns |
| 30-49 | Experimental |
| 0-29 | Speculative |
Why Deciduous?
Claude loses context. Sessions end, memory compacts, decisions evaporate. Six months later, no one remembers why you chose approach A over approach B.
Deciduous fixes this. Every decision is tracked in a queryable graph that persists forever. When Claude starts a new session, it can query the graph to understand past decisions - even from sessions it never saw.
This isn't documentation written after the fact. It's a real-time record of how software gets built, captured as decisions happen.
Building from Source
License
MIT