trust-line

Agentic Payment Trust Mapping — a Computational Prototyping tool that maps the Trust Line for autonomous payment feature decisions.

Input a payment feature scenario. Simulate how five user archetypes respond across the Identity Control ↔ Agent Autonomy spectrum. Receive an Intervention Boundary specification and design brief your engineers can act on directly.

Live demo: trust-line-theta.vercel.app

---

What it does

UX teams spend weeks trying to answer one question before shipping agentic features: when should the AI act silently, and when should it ask?

trust-line answers it in minutes.

For any autonomous payment feature you define, it simulates five behaviorally-grounded user archetypes:

Archetype	Trust tendency
Autonomous Delegator	Comfortable with agents acting for routine financial tasks
Convenience Maximiser	Prioritises frictionlessness; tolerates autonomy if errors are recoverable
Supervised Controller	Wants visibility above personal thresholds; trust earned incrementally
Privacy Sceptic	Wants explicit consent at each autonomous step regardless of amount
Risk Avoider	Treats all financial decisions as high-stakes; defaults to confirmation

For each archetype, the tool generates a three-tier Intervention Boundary:

SILENT AUTONOMY      — acts without notification when: …
NOTIFY AND PROCEED   — acts but notifies when: …
CONFIRM BEFORE ACTING — pauses for explicit confirmation when: …

This output format maps directly to what a UX Engineer or designer needs to spec agent interaction patterns.

---

Stack

• Next.js 14 (App Router, TypeScript strict mode)
• Gemini 3.1 Pro via @google/genai — server-side only, API key never touches the client bundle
• Tailwind CSS — dark navy / Google blue payments aesthetic
• Recharts — Trust Line spectrum visualisation
• Streaming responses — each archetype streams sequentially, Trust Line builds in real time

---

Local setup

# 1. Clone
git clone https://github.com/kgkazakos/trust-line.git
cd trust-line

# 2. Install dependencies
npm install

# 3. Add your Gemini API key
cp .env.local.example .env.local
# Edit .env.local and set GEMINI_API_KEY=your_key_here
# Get a key at: https://aistudio.google.com/app/apikey

# 4. Run
npm run dev

Open http://localhost:3000.

---

Deploying to Vercel

# Install Vercel CLI
npm i -g vercel

# Deploy
vercel

# Set the environment variable in the Vercel dashboard:
# Project Settings → Environment Variables → GEMINI_API_KEY

Or connect the GitHub repo directly in the Vercel dashboard and set GEMINI_API_KEY in project settings. The API key is consumed server-side only and never included in the client bundle.

---

Architecture

trust-line/
├── app/
│   ├── page.tsx                    # Main UI — scenario input, simulation state, results
│   ├── layout.tsx                  # Metadata, font imports
│   ├── globals.css                 # DM Mono + DM Sans, dark navy theme, trust gradient
│   └── api/simulate/route.ts       # Server-side Gemini API — streaming SSE, never client-exposed
├── components/
│   ├── ScenarioForm.tsx            # Payment feature input with validation
│   ├── ArchetypeSelector.tsx       # Archetype selection with colour-coded toggles
│   ├── TrustLineChart.tsx          # Horizontal spectrum with archetype markers
│   ├── ArchetypeCard.tsx           # Individual result with expandable boundary spec
│   ├── InterventionBoundary.tsx    # Three-tier Silent / Notify / Confirm display
│   ├── DesignBrief.tsx             # Copyable full spec in terminal style
│   └── LoadingState.tsx            # Progressive per-archetype streaming indicator
├── lib/
│   ├── prompts.ts                  # All prompts separated — zero inline strings
│   ├── archetypes.ts               # Archetype definitions, colours, default positions
│   └── types.ts                    # TypeScript interfaces — strict, zero any
└── .env.local.example              # GEMINI_API_KEY placeholder

Security:

• GEMINI_API_KEY is read only in app/api/simulate/route.ts (Node.js server runtime)
• No API key is included in the client bundle under any condition
• .env.local is in .gitignore

---

Part of the CPR Ecosystem

trust-line is a Computational Prototyping tool — part of the Computational Product Research (CPR) discipline developed by Kostas Kazakos.

Related projects:

• CausalTrack — Multimodal Say-Do Gap detection (Gemini + Neo4j), targeting CHI 2027
• Synthetic User Council — Agent-led ecosystem simulation, targeting CHI 2027
• CPR Orbital — 40 open-source research tools shipping every Friday, Mar–Dec 2026

---

License

MIT © 2026 Kostas Kazakos