afterbuild/ops
§ S-09/retainer-support

AI app retainer support — 40 hours, rules-file governance, priority rescue.

A Lovable support retainer that installs an agent rules file, reviews every Cursor and Claude Code PR before merge, and answers the phone at 10pm when the webhook stops firing. Month-to-month, 30-day cancel, no annual contract.

price
$3,499/mo
hours
40h
critical SLA
4h
Quick verdict

AI app retainer support is a $3,499/month rolling engagement: 40 hours of senior engineering on your Lovable, Bolt, Cursor, or Claude Code app, plus priority rescue queue with 4-hour business-day response on production-down incidents. Every retainer installs and maintains an agent rules file (.cursorrules / CLAUDE.md) so regressions caught in review cannot re-land. Includes monthly security scan, Slack access, and a written monthly changelog. Unused hours carry over 30 days. Month-to-month with 30-days' notice to cancel — no annual contracts.

§ 01/diagnostic matrix

Symptoms an AI app monthly retainer fixes

The Lovable support retainer and Cursor regression retainer share a signature: the agent keeps re-introducing the same bugs because no one is maintaining the guardrails between prompts.

symptom → root cause → what the retainer ships
SymptomRoot causeRetainer response
Cursor regenerates a whole file and breaks the surrounding moduleNo agent policy file; every prompt invents its own conventionsWe install and maintain `.cursorrules` / `CLAUDE.md` with stack rules, naming, and forbidden patterns
Claude Code keeps introducing the same three bugs across promptsNo regression-pattern memory; each session starts from scratchRules file maintains an anti-patterns section updated every time a regression is caught in review
Weekly Lovable prompt loop keeps undoing last week's fixLovable does not respect prior code review disciplinePolicy file pinned in project scope + weekly review of every Lovable-generated PR before merge
One-off freelancer ghosts on Friday at 10pmNo continuity; every incident starts by re-explaining the stackPriority rescue queue with 4-hour business-day response on production-down and a number to call
Founders burning weekends debugging the AI's outputNo senior review layer before the PR mergesSenior AI-PR review on every Cursor / Claude Code / Copilot / Windsurf PR with 24-hour turnaround
Monthly retainer invoices with no visibility into what got doneNo reporting cadence; hours vanish into a black boxWritten monthly changelog with shipped work, hours used, risks flagged, and next month's top three
§ 02/engagement schedule

Monthly Cursor regression retainer schedule

From kick-off onboarding to month-end changelog and rolling renewal. Rules file installed week one, review cadence live week two, priority rescue queue live week three.

  1. W1Kick-off

    Onboarding & rules-file install

    Onboarding call, repo/monitoring access, shared Slack. We install the first draft of the agent policy file (.cursorrules / CLAUDE.md) and agree the month's priority list.

  2. W2Cadence start

    Weekly AI-PR review begins

    Every Cursor / Claude Code / Copilot / Windsurf PR routes through senior review with 24-hour turnaround. Regressions caught feed straight back into the rules file.

  3. W3+Continuous

    Priority rescue queue live

    Production-down tickets jump the queue with 4-hour business-day response. Standard work ships on the agreed priority list. You approve every merge.

  4. M1Month-end

    Monthly changelog + rules-file diff

    Written changelog: hours used, what shipped, what's pending, risks. Plus the rules-file diff — every new rule added after a regression catch is documented.

  5. M+Renewal

    Rolling renewal or clean hand-off

    Month-to-month by default. Cancel anytime with 30 days' notice; we hand off clean with rules file, runbooks, and regression floor intact. Most clients stay 6–12 months.

§ 03/rules file

The agent rules file a retainer installs and maintains

This is the exact shape of the .cursorrules / CLAUDE.md we install in retainer week one. Every regression caught in review lands in the anti-regressions section so the same bug cannot ship twice. It is the single biggest lever on AI-app quality across Cursor, Claude Code, Copilot, and Windsurf.

.cursorrules + CLAUDE.md
markdown
01// .cursorrules  (installed W1, updated on every regression caught in review)02 03# Afterbuild Labs / Project Rules — maintained under Retainer Support04 05## Stack06- Next.js 16 App Router · React Server Components by default07- Supabase Postgres with Row-Level Security ON for every table08- Stripe via `stripe-node`; webhook handlers MUST verify signatures + be idempotent09 10## Forbidden patterns (auto-reject in review)11- `service_role` key referenced anywhere outside `/server/` handlers12- Client components importing from `/server/` boundary13- New tables created without a matching RLS policy in the same PR14- Stripe webhook handlers without `stripe.webhooks.constructEvent(…)`15- Raw SQL with template interpolation (use parameterised queries or Supabase query builder)16 17## Required patterns18- Every new route gets a matching Playwright smoke test in `tests/e2e/`19- Every payment handler has a Vitest idempotency test20- Env vars consumed via `import { env } from "@/env"` — never `process.env` directly21- Errors thrown as typed classes from `@/lib/errors`; logged via `logger`22 23## Anti-regressions (carry forward every month)24- 2026-02-14 · session cookies must include `sameSite: "lax"` + `secure: true`25- 2026-03-03 · `app/api/stripe/webhook/route.ts` is Node runtime, not edge26- 2026-03-21 · never call `revalidatePath` inside a loop — batch it
Installed in retainer week one. Maintained every month. The anti-regressions section grows with every review catch.
§ 04/monthly changelog

What the AI app monthly retainer changelog looks like

Sent the first business day of the next month. One page. No meeting required. Auditable for quarterly board reports.

Retainer changelog — March 2026
Hours
37.5 / 40 used · 2.5 carry-over
Shipped
  • — Stripe subscription-upgrade webhook handler (idempotent)
  • — Supabase RLS tightened on `invitations` table
  • — Next.js 16 upgrade + hydration regressions fixed
  • — 4 Cursor PRs reviewed · 1 blocked, 3 merged
Risks flagged
  • — 2 dependency CVEs (high) queued for April patch
  • — Redis memory at 72% — bumping next sprint
Rules-file diff
  • + 2026-03-03 webhook route = Node, not edge
  • + 2026-03-21 no `revalidatePath` in loops
Next month top 3
  1. 1. Ship invite-flow v2 (10h)
  2. 2. Supabase pooler migration (8h)
  3. 3. CVE patches + regression sweep (6h)
§ 05/deliverables

What the Lovable support retainer ships every month

§ 06/pricing

AI app retainer support — one tier, priced clean

standard
price
$3,499/mo
turnaround
40 hours / month
scope
Senior engineering on your Lovable, Bolt, Cursor, or Claude Code app. Rules-file governance, AI-PR review, priority rescue queue, monthly security scan, written changelog.
guarantee
Unused hours carry 30 days. Month-to-month, 30-day cancel. No annual contract.
Start retainer
Overflow & expanded tiers

Overflow hours billed at $175/hr. Expanded retainer with 80 hours/mo at $5,999 quoted on the onboarding call. On-call retainer with 24/7 PagerDuty quoted separately — typical shape adds a second engineer.

Who this is not for

If you need less than 25 hours consistently, run Ongoing Maintenance Starter at $1,500/mo instead. If you need pre-launch build work, start with Finish My MVP.

§ 07/vs alternatives

Retainer vs hourly vs full-time hire

DimensionHourly freelanceFull-time hireAfterbuild Labs retainer
Monthly cost$175-250/hr · unpredictable$250k+ fully loaded$3,499 fixed · 40h
Agent rules fileNot in scopeIf they rememberInstalled W1, maintained monthly
AI-PR review SLANo SLAWhenever free24 hours every PR
Critical SLANoneWorking hours4 business hours
Lock-inPer engagement12-24 months + severance30-day cancel

Pick AI app retainer support if…

  • You launched an AI-built app (Lovable, Bolt, Cursor, Claude Code) and now ship weekly with AI tools
  • You want a senior engineer reviewing every Cursor / Claude Code PR before merge
  • You need an AI app monthly retainer with a rules file keeping the agent inside the guardrails
  • You are a non-engineer founder running one engineer plus AI tools and need a fractional CTO-equivalent
  • You had one production incident too many with no one senior to call at 10pm
  • You are between full-time hires and need the bridge without $250k salary and equity

Don't pick this retainer if…

  • You need 24/7 on-call with pager rotations — we quote that separately as a Scale+ retainer
  • You need 80+ hours/month — we quote an expanded retainer ($5,999+) instead of stacking
  • You want a dev shop to own product roadmap (we own production, you own product)
  • You are pre-launch with no paying users — start with Finish My MVP instead
  • You want open-ended product strategy or sales help — that's not engineering hours
The rules file was the unlock. Cursor stopped reinventing conventions and Claude Code stopped shipping the same three bugs. Weekly merge velocity doubled inside a month.
CTO· B2B SaaS · Cursor regression retainer, month 2
FAQ
What counts as AI app retainer engineering hours?
Anything that moves the codebase or infrastructure forward: features, bug fixes, refactors, dependency upgrades, monitoring setup, security patches, agent rules-file maintenance, senior review of your in-house dev's PRs, and pair sessions. Does not count: open-ended product strategy, recruiting help, or sales calls. Every hour is logged against the retainer and reported in the month-end changelog.
What happens if I need more than 40 hours in a month?
We flag it by week 3 and offer either (a) overflow hours at $175/hr, or (b) a second retainer block for the month. No surprise invoices. If 40 hours is consistently too few, upgrade to a custom retainer ($5,999+ for 80 hours) or bring in a second engineer. Most Lovable support retainer clients find 40 hours lands right for an app with real traffic and one in-house dev shipping Cursor / Claude Code PRs.
What if I use less than 40 hours in the month?
Unused hours carry over for 30 days — they expire at the end of the next month, not indefinitely. Keeps us honest about capacity and stops you from paying for dead hours. If you are consistently under 25 hours, we will suggest Ongoing Maintenance Starter at $1,500/mo instead — no point overpaying.
Do you maintain an agent rules file for Cursor and Claude Code?
Yes — this is the core of the retainer. We install a .cursorrules and CLAUDE.md in week one with your stack conventions, forbidden patterns, and required patterns. Every regression we catch in review gets written back into the rules file so the same bug cannot land twice. By month three most retainers have a rules file that prevents 60-80% of the regressions the AI was shipping before we started.
What if my tool isn't Lovable, Bolt, Cursor, or Claude Code?
The Cursor regression retainer works for Lovable, Bolt, Cursor, Claude Code, Copilot, Windsurf, Replit, v0, Base44, and hand-written TypeScript or Python apps. If you're on something exotic (Rails, Django, Laravel, .NET) we'll scope on the onboarding call. The rules file and review cadence are the same shape; the stack specifics change.
Do you work on weekends and overnight?
Priority rescue queue covers production-down incidents 7 days a week with 4-hour business-day response and best-effort outside business hours. Non-emergency work is business-hours only. For 24/7 on-call, we quote a custom retainer with a separate on-call engineer carrying PagerDuty.
Can I pause the retainer instead of cancelling?
Yes — up to 2 months per 12-month period, 30 days' notice. Useful if you're mid-fundraise or off-season. We keep your priority slot and pause any unused carry-over. Rules file and runbooks stay live in the repo so nothing regresses during the pause.
What does the monthly changelog look like?
A one-page written summary sent the first business day of the next month: hours used (vs 40), ships (feature / bug / CVE), risks flagged, rules-file diff (new anti-regression entries added), and next month's top three priorities. Keeps everyone aligned without another meeting and is auditable for quarterly board reports.
40h
Senior engineering / month
24h
AI-PR review turnaround
Every Cursor / Claude Code PR
4h
Critical rescue SLA
30d
Cancel notice
No annual contract
§ 09/related scope

Related AI-app rescue services and retainer case studies

Next step

A senior engineer on tap. Rules file installed.

$3,499/mo. 40 hours. Priority rescue queue. Cancel any time with 30 days’ notice. Cheaper than one day of a full-time hire, every month.