← All docs · source: DEMO.md

LazyToken — Demo Environment

A repeatable, synthetic-data demo for sales conversations and evaluations. It stands up a LazyToken server populated with a realistic mid-size R&D org so every dashboard screen tells a story — without any real customer data.

The demo dataset is generated by server/scripts/seed-demo.ts (npm script seed:demo). It is deterministic (same numbers every run) and self-contained: it creates its own org and never touches the small seed.ts fixture used by tests.

1. What the demo represents #

Northwind Labs — a 60-seat, Enterprise-tier org, ~90 days into adoption:

Dimension	What the seed creates
Teams	Platform, Payments, Data, Mobile, SRE — each "goes live" on a different day, producing a real adoption curve
Developers	~22 developer accounts across the five teams
Devices	~21 active devices (macOS/Linux/Windows), 2 "silent" (stale), 1 revoked, one on an older agent version
AI agents	Copilot and Claude Code as first-class citizens (Copilot-heavy, mirroring a Microsoft shop), plus a Cursor / Gemini CLI tail — per PRD §3.7 agent-equality
Metrics	~90 days of daily savings with weekday seasonality; ~7–8k raw metric rows
Opportunities	`kubectl` / `docker` / `terraform` show as high-volume, low-savings passthrough (≈8–12% saved) next to build/test tools at 76–85%
Security events	Context Firewall (DLP) events across all rule types — `aws-key`, `github-token`, `openai-key`, `anthropic-key`, `private-key`, `connection-string`, `jwt`, `high-entropy`, `guardrail-block` — type + tool + time only, never a matched value
Licensing	A live Enterprise license window so the header shows a healthy, non-expired license

Roughly: ~54M raw tokens → ~28M after filtering (~49% saved org-wide), with the strong-savings tools far higher and passthrough tools deliberately low so the Opportunities story lands.

2. Launch the demo #

Assumes Node 22+ and PostgreSQL 16 (or use the Docker path from the Quick Start). Use a dedicated database so the demo is easy to reset:

cd server
npm ci

createdb lazytoken_demo
DATABASE_URL=postgresql://<user>@localhost:5432/lazytoken_demo npx prisma migrate deploy
DATABASE_URL=postgresql://<user>@localhost:5432/lazytoken_demo npm run seed:demo

# Start the server against the SAME database:
DATABASE_URL=postgresql://<user>@localhost:5432/lazytoken_demo npm run dev

The seed prints the logins and a live enroll token. Open the dashboard (default http://localhost:8080).

Reset / re-seed: the seed refuses to run twice into the same DB. To start over, drop and recreate the database:

dropdb lazytoken_demo && createdb lazytoken_demo
DATABASE_URL=...lazytoken_demo npx prisma migrate deploy
DATABASE_URL=...lazytoken_demo npm run seed:demo

3. Demo logins #

Role	Email	Password	Use it to show
Admin	`admin@northwind.demo`	`demo-admin-1234`	Everything — policies, users, licensing, settings
Finance	`finance@northwind.demo`	`demo-finance-1234`	The ROI / cost angle (CFO persona)
Lead	`lead@northwind.demo`	`demo-lead-1234`	Team-scoped view (Platform team lead)
Viewer	`viewer@northwind.demo`	`demo-viewer-1234`	Read-only executive view

These are synthetic demo credentials for a throwaway database. Never reuse them for anything real.

4. Suggested demo script (≈8 minutes) #

Open on Executive as admin@northwind.demo. Lead with the headline: "~49% of tokens filtered out before they ever reached the model, across ~90 days." Point at the adoption curve — teams coming online one after another.
Teams. Show the spread across Platform / Payments / Data / Mobile / SRE and drill into one. Mention the anonymization mode ("Dev #N") for works-councils and GDPR-sensitive orgs (see the Security Whitepaper).
Commands / Opportunities. Contrast the strong savers (git, cargo, pytest at ~80%) with kubectl / docker / terraform flagged as passthrough opportunities — "here's where we can tune filters next."
Security (the differentiator). Open the Context Firewall screen: counts by type over time — AWS keys, GitHub/OpenAI/Anthropic tokens, private keys, guardrail blocks. Emphasize: we log that a secret was blocked, never the secret itself. This is the moment the conversation moves to the security budget.
Fleet. Show active vs. silent devices and an out-of-date agent — the "central control" story for the DevEx/Platform persona.
Agent equality. Filter by AI agent to show Copilot and Claude Code side by side — "we're not a Claude-only tool; we cover the fleet you actually run."

5. What to say about the data #

It is synthetic — generated locally, deterministic, no customer involved.
The shape is realistic (adoption ramp, weekend dips, passthrough tools, DLP event mix) so the screens behave like production.
Nothing in the demo dataset violates the privacy model: metrics are numbers + tool names; security events are {type, tool, time} only. It is the same allowlist the real agent is bound to.

6. Notes and honest caveats #

The demo uses password logins for simplicity. SSO (OIDC/SAML) is configured per real org and is not part of the synthetic seed.
The dashboard reads only from the aggregated tables the rollup populates; the seed runs that rollup for you, so every aggregated screen is ready immediately.
USD values depend on the org's price table. The demo sets a per-agent price table (claude-code $3, copilot $2, cursor $3, gemini-cli $1.25 per MTok); adjust it under Settings → Token prices to match a prospect's real contract and watch the ROI number move live.