MemberIntel V1.5 Specification

Status: Draft v0.1 (2026-04-29)
Owner: Blair (CEO, MemberPress / Caseproof)
Working name: MemberIntel (internal — not committed)
Builds on: SPEC-v1.md (the canonical V1 spec)

---

1. Overview

MemberIntel V1.5 extends the V1 product (advisor + dashboard + two-tier brain) with two new capabilities that depend on each other technically:

1. AI Agent / Write-Actions. The AI can execute approved changes inside the customer’s MemberPress site (membership level CRUD, content protection rules, email / dunning configuration), with per-action confirmation, audit logging, and undo support.
2. “Set up FOR ME” Greenfield Wizard. A fast, opinionated, ~5-minute chat-driven onboarding that interviews a new MemberPress customer and uses the agent to set up their site for them — membership levels, content rules, starter email sequences. Surfaced inside MemberPress’s own onboarding flow as an opt-in “Set up FOR ME” choice; promoted via a 14-day Pro trial that aligns with MP’s existing 14-day money-back guarantee.

V1.5 is deliberately built on the V1 architecture without altering it. All V1 features, tier mechanics, brain architecture, data boundaries, compliance commitments, and pricing remain in force as defined in SPEC-v1.md. V1.5 introduces new features and a new trial mechanic; it does not change V1’s free tier, Pro tier, or core advisor experience.

The strategic intent of V1.5: turn MemberIntel from “advisor for existing operators” into a product that also sets sites up correctly from day one — driving MemberPress retention (better-configured sites churn less) while creating a clear, high-converting Pro upsell moment inside MP’s own onboarding funnel.

For unchanged scope (architecture, brain, compliance, base pricing model, base ICP), see SPEC-v1.md.

---

2. Goals

Primary

• Retention lift for new MP customers by getting their site correctly configured from day one via the “Set up FOR ME” wizard.
• Pro tier conversion by aligning a 14-day Pro trial with MP’s existing 14-day money-back guarantee window — customers experience full Pro value (agent + premium chat + advanced reports) during their MP MBG period.
• Deeper Pro value for existing customers by giving them an AI agent that can execute approved changes on their site, not just advise.

Secondary

• Build the foundation for V2 cross-platform support (Kajabi, Thinkific, PMP) by formalizing the agent surface against the MemberPress MCP, which other-platform connectors can mirror.
• Grow the brain (per-customer + global) faster by capturing real action data — what users tried, what worked.

V1.5 Success Metrics (priority order)

1. “Set up FOR ME” trial start rate — % of new MP customers who opt into the trial during their MP onboarding. Target: 25–40%.
2. Trial → Paid conversion rate — % of trial starters who convert to ongoing Pro on day 15. Target: 30–50% (higher than V1’s free-→-Pro conversion because card is required upfront and value is delivered fast).
3. Wizard completion rate — % of wizard starters who finish the interview and have their site set up. Target: 70%+.
4. Agent action success rate — % of agent-proposed actions that execute successfully without user-reported failures. Target: 98%+.
5. MP retention lift — change in 30-day MP retention for customers who used the wizard vs. those who used MP’s standard onboarding. Target: measurable positive lift; specific number set after V1 baseline data.

---

3. Non-Goals (V1.5 explicit exclusions)

Inherits all SPEC-v1 non-goals. V1.5 specifically does NOT do:

• No Pro+ tier introduction. All V1.5 features go into the existing $29/mo Pro tier. Pro+ is V2+.
• No automated WordPress / plugin installation. V1.5 assumes the customer already has WordPress hosted and walks them through installing the MP plugin from a zip file. No host-specific automation, no SSH/WP-CLI, no browser automation.
• No host-specific install instructions. Generic “install a plugin from zip” guidance only.
• No agent automation of Stripe/PayPal connection. Agent walks the user to the place where they connect a payment processor inside MP admin (deep links + step-by-step instructions); does not OAuth into payment providers on the user’s behalf.
• No V2 features. Cross-platform support, SOC 2, tiered Pro plans, and cross-customer benchmarks remain V2+.
• No agent operations beyond the four approved categories (Section 5.1). Member account modifications, content publishing, plugin/theme management, etc. are out of scope.
• No undo for irreversible operations. Some actions (member emails sent, payments captured) cannot be undone; agent must clearly mark these as irreversible at confirmation time and require stronger confirmation.

---

4. ICP Updates

V1 ICP (existing MemberPress operators) remains unchanged. V1.5 adds:

First-time MemberPress customer (new persona)

• Profile: just installed MemberPress, possibly during a 14-day MBG window. Wants to launch a membership site but doesn’t know where to start with tier structure, pricing, content gating, or technical configuration.
• Pain points: setup is intimidating; making wrong decisions early creates pain to undo later; doesn’t have time or expertise to research membership best practices.
• Acquisition channel: encountered in MemberPress’s own onboarding flow as an opt-in “Set up FOR ME” choice. Email campaign to abandoned MP installers as a recovery channel.
• Their job: “I just want my site set up. Ask me a few questions and configure it for me.”

---

5. Agent / Write-Actions

5.1 Approved Actions (V1.5 scope)

The AI agent can execute the following operations on the customer’s MemberPress site, all gated by per-action user confirmation:

Category	Operations	Risk profile
Membership level CRUD	Create / update / archive levels (name, price, billing period, trial config)	Medium-high. Live members on a level make changes risky.
Content protection rules	Create / update rules mapping content (posts, pages, custom post types) to membership levels	Medium. Misconfiguration can expose paid content to free users.
Email / dunning configuration	Configure transactional emails (welcome, renewal reminder, failed-payment recovery, dunning sequence)	Low-medium. Reversible. High-leverage for churn reduction.

Out of V1.5 agent scope (deferred):

• Member account creation / modification / deletion.
• Direct Stripe / PayPal API operations (V1.5 only guides the user to the connect screen in MP admin).
• Content creation / publishing / editing.
• Theme / plugin management.
• WordPress core settings.
• User role / permission changes.

5.2 Integration Path: MemberPress MCP

V1.5 uses MemberPress’s forthcoming MCP (Model Context Protocol) server as the integration surface for write operations. MemberIntel does not build a bespoke MP-side connector; it consumes MP’s MCP tools.

• MemberPress ships its MCP independently (planned for the month preceding V1.5 build kickoff).
• MemberIntel’s agent layer wraps the MCP tools with the additional UX MemberIntel owns: confirmation modals, audit logging, undo orchestration, eval coverage.
• If an MP MCP tool is added or changed, MemberIntel can incorporate it without an MP code-side push.

5.3 Confirmation Model

Per-action confirmation, server-enforced. Every write action follows the same flow:

1. AI proposes the action in chat: “I’d like to create a Pro level at $29/mo, billed monthly. Approve?”
2. UI presents a confirmation modal showing:
   • The exact change being made.
   • The expected impact (e.g., “0 existing members affected” or “12 members on Tier 2 will be migrated to Tier 3”).
   • Reversibility indicator: ✅ Reversible / ⚠️ Irreversible / 🔒 Strong-confirm required.
   • For irreversible / strong-confirm actions: a typed-confirmation field (e.g., type the level name to confirm).
3. User clicks Confirm or Cancel.
4. On Confirm: MemberIntel calls the MP MCP tool. Result returned. Audit log entry written.
5. On Cancel: action is logged as proposed-but-rejected for AI feedback.

No batching, no plan-level approval in V1.5. Each step requires its own confirm. Plan-level + edit-before-execute is a V2 candidate.

5.4 Audit Log

Every agent action (proposed, confirmed, executed, failed, undone) is logged to a per-customer immutable audit log with:

• Timestamp.
• Acting user (the human who confirmed).
• Action category + specific operation.
• Before-state snapshot (relevant fields).
• After-state result (or error message if failed).
• Undo availability (yes / no, with reason if no).

The audit log is visible to the customer in their Settings → Audit Log screen and to support staff. Logs are retained for the life of the account; data deletion (GDPR) wipes the log along with everything else.

5.5 Undo Semantics

• Reversible actions (most level edits, content protection rules, email config changes): one-click undo button in the audit log entry. Restores prior state via a follow-up MCP call.
• Irreversible actions (level archival when it had live members, sending emails to a list, etc.): no undo. Audit log entry shows “Irreversible — undo not available.” User is warned at confirmation time.
• Failed actions (MP rejects the write, network error, MCP tool error): rolled back automatically. Audit log shows the failure and resulting state. AI surfaces the error in chat with retry / alternate-approach options.
• Multi-step plans where step N fails: prior steps are not auto-rolled-back unless explicitly tagged as transactional. AI surfaces the partial-completion state and offers to undo executed steps individually.

5.6 Safety Bounds

Server-enforced constraints regardless of what the AI proposes:

• Cannot delete (only archive) any level with live members.
• Cannot modify pricing on a level with N>0 active subscribers without typed confirmation.
• Cannot send a marketing email to >100 members without typed confirmation.
• Rate-limited: max 20 agent actions per hour per customer in V1.5.
• All operations gated by the MP-side permissions of the connected user; agent inherits, never escalates.

5.7 Eval Suite

A dedicated eval suite for agent actions runs on every release before deployment:

• Fixed test cases per action category (create level, update price, archive with live members, etc.).
• Both happy paths and adversarial cases (AI proposes destructive action, expected to require strong confirmation; AI proposes invalid action, expected to refuse).
• Pass rate must be 100% on safety-critical evals; >95% on non-critical evals.
• Eval failures block release to production; require sign-off to override.

---

6. “Set up FOR ME” Greenfield Wizard

6.1 Purpose

A fast, opinionated, ~5-minute chat-driven interview that lets a first-time MemberPress customer have their site configured by AI without manual setup work.

6.2 Entry Points

In priority order:

1. Inside MemberPress’s own onboarding wizard. When MP’s onboarding asks “How would you like to set up your site?”, “Set up FOR ME (free 14-day Pro trial)” is one of the options, alongside MP’s existing manual onboarding. MP’s existing wizard remains the default path; MemberIntel is the opt-in alternative.
2. In MP admin (post-install, pre-setup). Customers who installed MP but haven’t configured anything see a “Set up FOR ME” banner inside MP admin until they either complete setup themselves or accept the offer.
3. Email campaign to abandoned MP installers. Marketing email to people who installed MP but never finished setup. Links to memberintel.com onboarding.
4. memberintel.com homepage. A prospect can run the wizard before they’ve installed MP — wizard guides them through “install MP from a zip file” first, then continues into setup.

6.3 Trial Mechanics (the upsell engine)

• Selecting “Set up FOR ME” starts a 14-day Pro trial. Card required upfront.
• The 14-day window aligns with MP’s 14-day money-back guarantee — customer’s downside risk on Pro is co-extensive with their MP risk.
• Day 12: reminder email (“your trial ends in 2 days”).
• Day 14: card is auto-billed unless the customer cancels.
• Day 15+ if cancelled or refunded: account drops to MemberIntel free tier per SPEC-v1 §5.1 (chat-capped, monthly snapshot dashboard, no agent).
• Existing MP customers (already past their MBG window): same 14-day Pro trial offered as a one-time launch promo for the first 6 months of V1.5 availability. Card still required.

6.4 Wizard Flow

Fast and opinionated. ~5 minutes, 5–7 questions. Conversational chat UI.

Phase 1 — Interview (~3 minutes)

Branched chat that adapts based on answers. Core questions:

1. What are you selling or teaching? (free-text)
2. Who is it for? (free-text — audience description)
3. What format will your content be? (multi-select: video, written, community, downloads, live calls, other)
4. What’s your target monthly price point per member? (range: <$10, $10-25, $25-50, $50-100, $100+, custom)
5. How many tiers do you want at launch? (1, 2, 3, “let MemberIntel decide”)
6. Do you have a payment processor connected? (Stripe / PayPal / Neither — agent guides setup if Neither)
7. Anything else important about your business? (free-text, optional)

The wizard adapts: if the user picks “let MemberIntel decide” for tiers, AI suggests a structure based on the format + price + audience answers.

Phase 2 — Plan Preview (~30 seconds)

AI presents a one-screen summary of the proposed setup:

• Recommended membership levels with names + prices + billing periods.
• Recommended content gating strategy.
• Recommended starter email sequence (welcome, payment-failed recovery).
• Payment processor status (connected / needs connection).

User can accept, request changes (“make Tier 2 cheaper”), or save as draft and continue manually.

Phase 3 — Setup Execution (~1-2 minutes)

Agent executes the approved plan one step at a time. Each step shows a per-action confirmation modal (Section 5.3). User can step through quickly by hitting “Confirm” on each (similar to a wizard) or pause to review.

Phase 4 — Next Steps (post-setup)

After the agent completes setup, MemberIntel hands off with:

• A list of next things to do (publish your first content, configure email, customize copy).
• An invitation to ask the AI anything as questions come up.
• A link to the standard MemberIntel dashboard.

6.5 Non-Wizard Customer Path

Customers who go through MP’s standard onboarding (not “Set up FOR ME”) still see a persistent “Want help? Try MemberIntel” banner in MP admin. They can opt into a Pro trial later. Wizard is offered to them as part of that opt-in.

6.6 Plugin Install Help (for prospects who haven’t installed MP yet)

• AI provides generic plugin-install instructions: download MP zip → WP admin → Plugins → Add New → Upload Plugin → activate → enter license key.
• Single set of instructions, not host-specific. If the user’s host has unusual restrictions, they’re directed to MP support.
• After install + activation, the wizard resumes from Phase 1.

---

7. Pricing & Tier Updates

7.1 No Tier Restructure

V1.5 keeps the existing tier structure from SPEC-v1 §5: Free (Connected Free) and Pro ($29/mo). No Pro+, no annual plan changes, no agency tier in V1.5. Those are V2+ candidates.

7.2 V1.5 Trial Mechanic Replaces Pro-Only Trial

SPEC-v1 §5.2 defined a 14-day Pro free trial only for “direct-to-Pro signups via memberintel.com without an MP install” — primarily edge-case marketing flow.

V1.5 generalizes this: the 14-day Pro trial with card required is the standard upgrade path for any customer running the “Set up FOR ME” wizard, regardless of how they entered. This becomes the dominant trial path in V1.5.

The free tier signup path (no trial, just sign up free) remains available in parallel for customers who don’t go through the wizard.

7.3 V1.5 Feature Gating

Feature	Free	Pro / Pro Trial
All V1 features (advisor, dashboard, brain, etc.)	per SPEC-v1 §5.1	per SPEC-v1 §5.1
”Set up FOR ME” wizard	Not available	Available (start triggers the 14-day Pro trial)
Agent / write-actions on existing site	Not available	Available
Audit log	Read-only of past actions if a Pro→Free downgrade happened	Full read
Multi-step setup plans	Not available	Available

---

8. User Stories (V1.5-specific)

V1 user stories from SPEC-v1 §7 carry forward unchanged. New for V1.5:

8.1 Greenfield wizard

• US-V1.5-G1. I just installed MemberPress. MP’s onboarding asks “how do you want to set up?” I click “Set up FOR ME (free 14-day Pro trial).” I’m asked for a card up front. I enter it.
• US-V1.5-G2. I’m interviewed by an AI in chat for ~5 minutes about what I’m selling, my audience, my pricing, my tiers.
• US-V1.5-G3. I see a one-screen plan preview showing the levels, content rules, and emails the AI will set up. I approve.
• US-V1.5-G4. Agent walks through each setup step with a one-click confirm modal. I’m done in another minute or two.
• US-V1.5-G5. I see “Connect your Stripe to start accepting payments” with a deep link into MP admin’s payment connection screen.
• US-V1.5-G6. Day 12 of trial: I get a reminder email. Day 14: my card is charged. I’m now an ongoing Pro customer.
• US-V1.5-G7. Alternative: day 13 I cancel. Day 15: my account drops to MemberIntel free tier. My MP site keeps everything the wizard set up; only the ongoing MemberIntel Pro features go away.

8.2 Existing-Pro agent

• US-V1.5-A1. As a Pro customer, I ask the AI in chat: “Add a $99/month annual VIP tier with 2 months free.” AI confirms specifics, proposes the level + Stripe pricing recommendation, I confirm, agent executes. Audit log entry created.
• US-V1.5-A2. As a Pro customer, the dashboard surfaces an insight card: “Your dunning sequence is missing a day-7 follow-up. Add it?” One click → agent proposes the email, I confirm wording, it’s saved.
• US-V1.5-A3. I made a mistake — agent created a level with the wrong price. I open Settings → Audit Log, find the entry, click Undo. The level is reverted.

8.3 First-time creator (pre-MP-install)

• US-V1.5-N1. I land on memberintel.com from a Google search. I click “Plan and set up my membership site.” I’m interviewed by AI without signing in.
• US-V1.5-N2. AI generates a recommended setup plan as a document. I’m prompted: “to actually set this up, you’ll need MemberPress installed on a WordPress site. Here’s how.” Generic plugin-install instructions follow.
• US-V1.5-N3. Once I’ve installed and activated MP with my license, I return to the wizard and it executes the plan against my new MP site (Pro trial begins, card required).

---

9. UI Updates

V1 UI per SPEC-v1 §9 carries forward. New / changed for V1.5:

• Wizard surface. Full-screen chat-style flow inside MemberIntel, embeddable inside MP admin via iframe or shared-session redirect. Mobile-friendly.
• Confirmation modal. Standardized component used by every agent action (Section 5.3). One UI, consistent everywhere.
• Audit Log screen. New screen in Settings. Filterable by date, action category, status. Each entry shows full before/after, plus an Undo button when reversible.
• In-chat agent action cards. When the AI proposes an action mid-conversation, it renders a compact action card inline (action summary + Confirm / Cancel) rather than only the modal. Modal opens on confirm for the final review.
• MP admin banner (“Want help? Try MemberIntel”). New banner shown in MP admin to customers who didn’t take the wizard at signup. Persistent (dismissible) until they either accept or pass the MBG window.
• Pricing page update at memberintel.com. Reflect the new wizard offering and trial mechanics.

---

10. Compliance & Privacy Updates

V1 compliance commitments per SPEC-v1 §10 carry forward unchanged. Additions for V1.5:

• Agent ToS clause. New ToS section governs the agent: explicit consent for AI to execute writes against the user’s MP site, scope of authorized actions, indemnification language for outcomes the user confirmed.
• Audit log retention. Audit log is retained for the life of the account; explicitly removed on data deletion. Disclosed in Privacy Policy.
• Trial billing disclosure. Card-required trial mechanics fully disclosed at signup: trial length, charge date, cancellation path. Comply with applicable subscription-disclosure rules (e.g., FTC ROSCA in the US, similar elsewhere).
• Sensitive operation logging. Any agent operation flagged as irreversible or high-impact is additionally logged with the AI’s full reasoning context for post-hoc review if needed.

---

11. Technical Stack Updates

V1 stack per SPEC-v1 §12 carries forward. V1.5 additions:

• MP MCP integration. MemberIntel agent layer consumes the MemberPress MCP. Engineering owns the wrapper logic (confirmation, audit, undo orchestration). MP team owns the MCP itself.
• Audit log store. New Postgres table per tenant. Append-only design. Schema versioned.
• Undo orchestration service. Lightweight service that knows how to reverse each operation type. Stores enough before-state to enable rollback.
• Eval harness. Test suite that runs against a staging MP instance; runs on every PR + every deploy to validate agent action correctness.
• Trial state machine. Extends the entitlement layer (per SPEC-v1 §12) to track trial start, day-12 reminder fired, day-14 charge attempted, charge succeeded / failed / cancelled. Drives feature gating + comms.

---

12. Implementation Phases

Assumes V1.5 build kicks off immediately after V1 launch with the same team: 4 engineers + 1 designer + 1 PM + 1 content lead. Estimated V1.5 duration: 5–6 months.

Phase 1 — Agent Foundation (Months 1-2)

• MP MCP integration + wrapper layer.
• Confirmation modal component + UX patterns.
• Audit log store + UI screen.
• Undo orchestration for the three agent action categories.
• Eval harness + initial eval suite for membership level CRUD.
• Server-enforced safety bounds (rate limits, typed confirmation for irreversible).
• Internal testing against a staging MP instance.
• Milestone (Month 2): internal team can use the agent end-to-end on test sites.

Phase 2 — Greenfield Wizard (Months 2-4)

• Wizard UI (full-screen chat flow + plan preview screen).
• Wizard prompt design + plan-generation logic.
• MP admin banner + MP onboarding integration (“Set up FOR ME” option).
• 14-day Pro trial state machine (card upfront, day-12 reminder, day-14 charge, downgrade flow).
• Eval coverage for content protection + email/dunning agent actions.
• Email campaign for abandoned MP installers.
• Plugin install instruction component (memberintel.com path for prospects without MP yet).
• Milestone (Month 4): beta with 20-50 first-time MP customers; measure trial start, completion, conversion.

Phase 3 — Polish + Launch (Months 4-6)

• Iterate on wizard based on beta data (likely tuning prompts, plan defaults, reduction or expansion of question count).
• Iterate on confirmation UX based on agent action data.
• Brain expansion: content lead writes additional 30+ playbooks tuned for first-time-creator scenarios.
• Pricing page update at memberintel.com.
• Public launch.

V2+ (post-V1.5)

• Plan-level confirmation + edit-before-execute UX.
• Pro+ tier introduction (price ~$79-99/mo, additional features TBD).
• Cross-platform support (Kajabi, Thinkific, Paid Memberships Pro).
• SOC 2 Type 1.
• Cross-customer benchmarks (if legal Phase 0 work is unblocked).
• Expanded agent action surface (member account ops, content publishing, etc.).
• Agency tier.

---

13. Risks & Open Questions

Risks specific to V1.5

1. Agent action error blast radius. A bug in the agent or MCP that mis-executes a customer’s site setup is a brand-defining failure. Mitigation: 100%-pass eval suite for safety-critical cases as a release gate; staged rollout (alpha → beta → general); limited action surface in V1.5; per-action confirmation always.
2. Trial conversion underperforms. If trial → paid conversion lands at <20%, the unit economics of providing a fully-staffed wizard for free for 14 days don’t work. Mitigation: card required upfront filters for serious users; rapid iteration on plan quality; A/B testing of wizard length and depth.
3. MP MCP delivery risk. V1.5 depends on the MemberPress MCP shipping in usable form before V1.5 build kicks off. If the MCP slips, V1.5 has to either wait or build a bespoke connector (adds ~2 months).
4. MP onboarding integration friction. “Set up FOR ME” lives inside MP’s own onboarding wizard. Requires MP team alignment + UI changes on the MP side. Risk of cross-team coordination delays; mitigation: scope the MP-side work narrowly (one new option in their wizard with a redirect URL).
5. Wizard misconfiguration. AI generates a plan that sets up the customer’s site in a way that doesn’t fit their actual business; customer is dissatisfied even though the agent executed correctly. Mitigation: plan preview + edit step; “set up draft, customer can refine” framing rather than “we got it perfect first try.”
6. Undo expectations. Customers expect undo for everything; some operations (sent emails, captured payments, deleted data) genuinely cannot be undone. Mitigation: clear UI labels at confirmation time; conservative defaults (e.g., emails always queued for review by default in V1.5, send is a separate explicit action).
7. Brain quality dilution from wizard plans. Wizard interactions are high-volume + relatively shallow per session. Cross-pollination signal weighting (per SPEC-v1 §6.3) needs to handle this — wizard plans alone shouldn’t drive global brain content; only post-wizard ongoing engagement should.
8. Trial-end downgrade UX. Customers who started trial, didn’t convert, end up dropped to free tier on day 15 may feel surprised even with reminders. Mitigation: prominent in-app messaging during trial; clear day-15 email explaining what changed.

Risks unchanged from SPEC-v1

Differentiation from generic AI; hiring; site data sync reliability; hallucination; content lead bottleneck; cross-pollination privacy slip; support burden; brand dilution.

Open Questions

• Q1. MP MCP scope — confirm exactly which write tools MP’s MCP will expose by V1.5 build kickoff. The agent action categories in this spec assume create/update/archive on levels, content rules, and email config are all available.
• Q2. MP onboarding integration UX — final placement and copy for “Set up FOR ME” inside MP’s own wizard. Owned by MP product team, MemberIntel team consults.
• Q3. Card-required trial activation rate — during Phase 0 of V1.5, run a customer survey or analyze MP signup data to estimate what % will enter card upfront. Affects V1.5 success metric targets.
• Q4. Eval suite ownership — single engineer or rotating? Eval coverage standards (which actions are “safety critical” requiring 100% pass)?
• Q5. Audit log retention regulation — confirm with counsel that “life of account” retention satisfies GDPR / CCPA requirements when paired with a “delete account → wipe log” mechanism.
• Q6. Wizard fallback if MP MCP throws an error mid-setup — does the wizard retry, surface to user, mark plan as failed? UX needs to be designed for partial-completion states.
• Q7. Re-running the wizard — can a customer re-run “Set up FOR ME” after their initial run (e.g., 3 months later they want to restructure)? In V1.5, probably gated to Pro and requires a “you have an existing setup; this will modify it” confirmation. Worth designing.
• Q8. Existing MP customer trial promo — the one-time 14-day Pro trial offer for existing MP customers (post-MBG). When does it run? What’s the eligibility window? How is it communicated?
• Q9. Agent rate limits — “20 agent actions per hour” is a starting number; tune based on real usage patterns in beta.
• Q10. Content lead workload — wizard launches will produce a flood of new per-customer brain content. Verify the content lead’s cross-pollination review queue scales; consider tooling improvements (filter by source, batch-review patterns).

---

14. Glossary (additions to SPEC-v1 §15)

• Agent / Write-Actions — the V1.5 capability allowing the AI to execute approved changes on the customer’s MemberPress site via the MP MCP, gated by per-action user confirmation.
• MP MCP — MemberPress’s Model Context Protocol server, which exposes structured tools for reading and writing to a MemberPress install. MemberIntel’s agent layer consumes this.
• “Set up FOR ME” Wizard — the V1.5 greenfield onboarding flow: chat-based interview + plan generation + agent-executed setup.
• Per-action confirmation — the V1.5 safety pattern in which every individual agent write requires its own user approval. No batching, no plan-level approval.
• Audit Log — per-customer immutable record of every agent action proposed, confirmed, executed, failed, or undone.
• Undo orchestration — service responsible for reversing reversible agent operations on demand.
• Trial state machine — entitlement-layer extension that tracks 14-day Pro trial lifecycle (start, reminders, charge, conversion or downgrade).