reference
MemberIntel — One-Page Executive Summary
The whole product on one page — what it is, how it's priced, what V1/V1.5/V2 ship, where the moat is, and what's deliberately out of scope. Written for Blair and Santiago to read in two minutes; everything else in the KB is the long-form backup.
For: Blair, Santiago — and anyone who needs the whole product on one page before opening a spec.
Source of truth: V1 spec, V1.5 spec, V2 spec. If this page contradicts those, the specs win.
Status: Draft. Built from the V1 → V2 specs Seth is basing decisions on.
What it is
An AI advisor for membership-site operators, sold as paid SaaS on top of MemberPress. Each customer gets a private AI that learns their site, plus a shared playbook (“global brain”) built from anonymized patterns across the whole customer base. Free tier captures adoption; Pro at $29/mo captures revenue.
Pricing (single model across V1, V1.5, V2)
- Free — capped chat on Anthropic Haiku, basic dashboard, weekly digest, brain contribution. Cheap to run; exists to grow the brain.
- Pro — $29/mo — unlimited Sonnet chat, live insights, advanced reports, brain editing. Single tier; no Pro+ until V3.
- V1.5 adds a 14-day Pro trial bolted onto MemberPress’s existing 14-day money-back guarantee.
- The conversion engine: free → Pro at the moment of felt value (chat cap hit, advanced report requested, wizard finishes).
The moat (read this if nothing else)
Two brains, phased in over three versions. A global brain curated by Caseproof staff and seeded from MemberPress KB content ships at V1 day one. A per-customer brain that learns one site over time arrives in V1.5, isolated by tenant_id with Row-Level Security. A cross-pollination pipeline that promotes anonymized patterns from per-customer brains into the global brain (with k-anonymity floors and permanent exclusion for sensitive verticals like medical and advocacy) activates at V2+ after content-lead review, k-anonymity checks, and opt-in consent — see ADR-0014 for the security boundary.
The longer a customer uses MemberIntel, the more their per-customer brain is worth (V1.5 onward). The more customers use it, the better the global brain gets for everyone (V2+ onward). This is a product-mechanics moat — it doesn’t require legal cover for pre-existing data, and it’s hard to copy without time and scale.
Phasing — what each version ships
| Version | What it adds | Who it’s for | Why it ships when it does |
|---|---|---|---|
| V1 (June build → Oct launch) | Advisor + dashboard + two-tier brain. Read-only — generates instructions, customer executes manually. | Existing MemberPress operators. | Smallest ship that proves the brain works and the funnel converts. No legal Phase 0 dependency. |
| V1.5 | AI agent / write-actions (4 categories: membership levels, content rules, email/dunning) via the MemberPress MCP, with confirm + audit + undo. “Set up FOR ME” greenfield wizard for first-time MP customers, ~5 minutes, surfaced inside MP’s own onboarding. | New MP customers + existing Pro customers who want more than advice. | Retention lift on new MP customers; high-converting Pro upsell inside MP’s MBG window. |
| V2 | BuddyBoss support (BB Memberships runs on MP under the hood — same data shape, sister-company channel). Full feature parity including agent + wizard. Introduces platform-tagging in the brain so V2.1 (PMP) and V2.2 (Kajabi) are incremental. | BuddyBoss community-membership operators. | TAM expansion with low engineering risk; sets up the real cross-platform abstraction for V2.1+. |
Architecture in one paragraph
Python 3.12 + FastAPI on GCP. pgvector on Postgres for retrieval; per-tenant brain isolated via Row Level Security. Anthropic Sonnet for Pro chat; Anthropic Haiku model for Free users. All model routing is server-side and never user-configurable — one llm.call wrapper enforces routing, entitlements, and cost caps. Auth via Google OAuth → JWT. Data sources V1: MemberPress site + Stripe. Cost discipline is a first-class constraint (free tier must not bleed money at scale).
Deliberately out of scope (so nobody re-litigates these)
- No cross-customer benchmarks built from Caseproof’s pre-existing Stripe Connect / MP telemetry data in V1. (Substitute moat = the two brains.)
- No third-party paid data (SEMrush, Ahrefs, etc.).
- No fine-tuning of model weights — all “learning” is brain content + retrieval.
- No SOC 2 in V1/V1.5/V2 — GDPR + CCPA only.
- No mobile app. Responsive web only.
- No Pro+ tier, no agency/per-seat pricing, until V3.
- No agent actions beyond the four approved V1.5 categories.
- No host-specific install automation (V1.5 walks customers to install MP, doesn’t SSH in).
Top risks Blair and Santiago should know about
- Free-tier cost economics. If Free runs on Sonnet by accident or routing fails, unit economics break. Mitigation: server-side routing, hard quotas, the
llm.callwrapper. - Anthropic vendor concentration. Pro tier depends on Sonnet. ADR-0005 tracks the mitigation plan.
- Cross-pollination privacy. The global brain only works if customers trust that their data doesn’t leak. K-anonymity floors + permanent exclusion for sensitive categories. This is also the active question Allen (Blair’s privacy counsel) is reviewing in the late-May session — see counsel work.
- V1.5 agent safety. Write-actions in a customer’s live site. Confirm-per-action, audit log, undo where reversible, hard refusal where not. Eval suite is must-build, not nice-to-have.
Success metrics — what we’re tracking
- V1: ≥10% of MP customers connect to MemberIntel in 6 months. 5–10% Free → Pro in 60 days. 70%+ Pro WAU. 200+ approved global-brain entries by month 6.
- V1.5: 25–40% of new MP customers opt into the “Set up FOR ME” trial. 30–50% trial → paid. 70%+ wizard completion. 98%+ agent action success.
- V2: 1,000+ BuddyBoss customers connected within 6 months of V2 launch. AI quality parity ≥90% on the eval suite vs. MP scenarios.
If you want to go deeper: start with V1 spec §1–§6 for the product, arch-overview for the engineering shape, and Blair’s decisions log for the open calls. Everything else in the KB is supporting detail behind these.