The build map. What's shipped, what's next, and a living record of the work.
The Core: NowPage is a live proof weapon. You sit across from someone, demonstrate value in real-time, and the page IS the demo. No pitch deck. No follow-up email. The page speaks for itself.
That promise is now delivered. Reveal Live ships the "3 taps to QR" moment — record, name, generate. The Credential Vault gives every team member their own API keys. Teams + RBAC means Derek and Will manage their own domains independently. The speed pipeline renders hero content in seconds while deeper pages build in the background.
The dependency chain held. Vault enabled team secrets. Teams enabled independent operation. Both enabled mobile BYOK transcription. Each phase shipped independently and delivered value on its own — exactly as designed.
Next: Three publishing surfaces. The CLI (Claude Code calling the API directly), the app (desktop and mobile Reveal), and a hosted MCP server that lets any chat interface — Claude.ai, ChatGPT, Perplexity — publish live pages from inside the conversation. You don't have to adopt a new tool. Your existing chat becomes the tool.
Why this was first: Every API key lived in Vercel env vars that only Jason could touch. Derek wanted to run Reveal with his own Anthropic key. Will wanted to test with a different model. They couldn't — bottlenecked on one person's credentials.
The Vault fixed this with AES-256-GCM encryption, full CRUD API, VaultManager UI, and resolveKey() integrated into all 6 studio routes. User vault → team vault → env var fallback. Simple, clean, done.
Shipped March 6, 2026.
Internal first, product later. Derek, Will, and Jason needed to share domains without stepping on each other's pages. That's now live.
Teams shipped with admin/editor/viewer roles, M:N domain access via a junction table (team_id OR user_id grants), MFA via Google Authenticator TOTP, audit logging for login/publish/move events, and an email allowlist for access control. Derek and Will independently manage their domains, store their own API keys, and see only what's relevant to them.
Shipped March 7–8, 2026.
This is the distribution play. A hosted MCP server (Vercel or VPS) registered with Claude.ai, ChatGPT, Perplexity, and any MCP-compatible interface. Users keep using whatever chat they already use — it gains the ability to publish live NowPage pages from inside the conversation.
Three publishing surfaces, one auth model:
CLI — Claude Code calls NowPage API directly. Power users, developers.
App — Desktop Reveal or mobile wizard. Full studio, profiles, analytics.
MCP — Any chat interface. User enters their API key once in settings. From that point, "publish this as a page on my domain" works from any conversation.
MCP tools exposed: create_page, publish_page, list_templates, get_page, improve_page. Each scoped to the user's role and domain permissions via their API key. The vault, team model, and RLS policies built in Phases 1–2 enforce every boundary.
The /publish skill makes the CLI a one-command flow. MCP makes the entire platform programmable — list pages, move them between domains, check analytics, read vault secrets — from any AI tool that supports MCP.
New role: Creator. Below viewer. Can create and publish to one page on a shared domain. Can't see other pages, can't manage teams, can't access the full studio. This is the $50/mo "room access" tier — MCP access + one live page. The entry point.
$50/mo — Creator: 1 page, MCP access, shared domain
$297/mo — Editor: Unlimited pages, own domain, studio profiles
$797/mo — Admin: Team management, multiple domains, full analytics
$2,497/mo — White-label: Custom branding, client-facing
This was the headline demo — and it's delivered. Reveal Live shipped Layers 10–13: live transcription via Deepgram Nova-3, the mobile wizard with responsive 3-tap flow, LinkedIn intel extraction (screenshot-to-text + URL parsing), the speed pipeline (complexity router + assembler for hero-first rendering), QR codes with branded placeholder pages, agent swarm (LinkedIn/Company/Social running in parallel), quality gates (factual auditor + steelman scoring), voice-to-improve (speak an instruction → targeted edits), live presets (configure once on desktop, use on mobile), and progressive rendering (content JSON preview while Pass 2 builds).
The "3 taps to QR" promise is real. Before an event, preset your template, CTA, and domain. At the event: record, name, generate. Show the QR code while they're still talking to you.
Shipped March 8, 2026.
30 features live in production across engine, studio, profiles, security, analytics, and Reveal Live.
Features that emerged from the primary phases — some shipped, some scoped for later.
Do the research once, use it everywhere. Three improvements to Reveal Live's research input:
Smart input — paste full URL, @username, or just a name. LLM resolves to the right profile URL.
Multi-URL — add as many links as needed per prospect. LinkedIn personal + company + podcast + blog + X. All routed to the right agents.
Camera capture — snap a photo of their profile on screen. Vision LLM extracts the URL and username automatically.
ICP Contact Library: After every Reveal run, the psychographic deep-dive (dreams, fears, suspicions, communication style, trigger phrases, DO/DON'T rules) is saved as a persistent contact profile. The library compounds — first run for a contact is full research (~30-60s). Follow-up runs, email funnels, proposals, and future templates use the cached profile instantly.
Hierarchical URLs with meaning. Pages already support domain/group/page/subpage via parent chains. What's missing: group pages that are useful without custom content.
A group page auto-renders from its children's metadata — principle statement at top, ordered child list below. Not a table of contents. A first-principles container that explains why these pages belong together, then lists the what. Optionally password-gated. Optionally a neural registry.
Focused neural registries are the real unlock. Instead of one domain-wide registry, each group scopes its own context. MCP tools requesting prds/vault automatically receive the /prds/ group registry — focused, relevant context injection without the noise.
Replace the basic list view with a sortable, searchable data table. Columns: name, path, folder, status, tags, updated date, actions. Search filters on name/slug/path in real time. Hierarchy indent shows parent/child relationships inline.
Add "+ Group" flow for creating page groups with auto-rendered index pages. Three URL structures supported: domain/page, domain/group/page, domain/group/page/subpage.
Three modes replace the Live/Desktop split. The real axis is time pressure, not device. A desktop user screen-sharing a Zoom call needs the same speed as mobile at a conference.
Realtime — 30-90 seconds. Active Zoom transcription or event mic. Speed pipeline auto-engaged. Desktop + mobile.
Studio — 2-5 minutes. Post-meeting builds with full depth. All templates available.
Engine — Background/scheduled. Intelligence briefs, daily content, newsletters. Batch processing.
13 unified categories replace the current 6. Playbooks merge into Lead Magnets (same raw template, 3 strategic CTAs). Sales & Presentations becomes first-class. Intelligence, Content Engine, LinkedIn, Newsletter, Email Sequences, and Client Deliverables are new. Each mode filters which categories are visible.
Saved production configurations. Every run currently requires setting 9+ variables: domain, voice, design, ICP, template, depth, gate, folder, input mode. Quickstarts capture all of these so repeat production work is one click.
"Mastery Minute — MasteryMade" pre-sets everything: masterymade.com domain, casual voice, MasteryMade design system, concise depth, no gate. Drop in today's insight, hit Go. "Athio JV Cold" pre-sets athio.ai, professional voice, password gate, URL input mode. Paste LinkedIn URL, hit Go.
Two creation paths: dedicated "New QS" wizard, or "Save as Quickstart" after manually configuring a run. Each QS shows average cost per run from usage history. Shareable with team members.
Self-service infrastructure. Managing Vercel env vars and running Supabase migrations requires switching to external dashboards that only Jason can access. This creates a bottleneck.
Three layers: Vercel env var sync (read/write/push-pull with vault), SQL editor with read-only mode, and migration runner with schema browser. All admin-only with full audit logging.
Every template now shows estimated cost before generation. Collapsible 2-column layout displays page counts, model costs, and total estimates. Users see exactly what a run will cost before they commit to it.
Full cost tracking infrastructure: per-run token counts, per-user spend aggregation, team admin cost breakdowns. Usage dashboard with daily charts and by-service/by-model views. MODEL_PRICING table covers every model in the pipeline.
Live demo artifacts and partner pages. The page IS the pitch.
A living record. What shipped, when, and what it means.