Plansmith
Curated planners for AI coding agents

A spec, forged from what actually shipped.

Point Claude Code, Cursor, or Codex at a PlanSmith planner. It interrogates you with cited checklists, then writes a complete, executable spec. Every line is grounded in hundreds of real, in-market products — 50+ closely reviewed per vertical, mined by hand, not imagined by a prompt.

100sapps surveyed across the catalog
50+closely reviewed per planner
100%checklist items cited
Specimen
B2B SaaS · flagship

SaaS Dashboard

feature inventory · 31 dashboards tabulated

Transactional email94%
Stripe as primary processor90%
Email+password auth / invite by email87%
3-tier RBAC (owner/admin/member)77%
Webhooks (outgoing)65%
Ranked by prevalence — 0 fabricated claims
Works with
Claude CodeCursorCodex CLIOpenAI CodexWindsurfAiderCline
The inventory · §01

The moat isn't the format. It's the ledger behind it.

For every vertical we survey hundreds of real products and closely review 50+, tabulating each feature into a frequency-ranked inventory. Your own AI can't fabricate these counts without redoing the research. Below is a slice of the SaaS Dashboard inventory — 31 sources tabulated.

Feature
Prevalence across 31 tabulated dashboards
Sources
Freq.
Transactional emailResend dominant in modern stacks; use unless >100k emails/mo
29/31
94%
Stripe as primary processorUniversal; wait for 2nd customer before adding Lemonsqueezy/Paddle/Polar
28/31
90%
Email+password auth / invite by emailTable stakes for any team product; argon2/bcrypt hashing standard in modern starters
27/31
87%
3-tier RBAC (owner/admin/member)The default everywhere; add tiers only when a customer explicitly asks
24/31
77%
Webhooks (outgoing)Required for any product with events; ship with signing secrets from day one
20/31
65%
Tiered freemium to proDefault for self-serve; skip if sales-led
20/31
65%
Usage-based / metered billingWire event tracking from day 1 — retrofitting metering is painful
13/31
42%
Audit log (admin-facing)Write events to table from v1; surface UI in v1.5+
11/31
35%
Saved views / favoritesBuild when first power-user complains; L1, L5, L18 ship it
8/31
26%

Three source layers per inventory: OSS starters · live SaaS · marketplace templates. The full ledger ships inside every planner — so your agent proposes the 87% feature before the 32% one, and tells you which is which.

Open the full inventory
The shop · §02

Ten verticals. Each one mined, not guessed.

Best seller
SKU·SAA-01SaaS Dashboard

SaaS Dashboard Planner

Turn 31 real SaaS dashboards into an executable CLAUDE.md — every checklist item cited, no architecture guesswork.

31 sources · frequency-ranked
$49View planner →
SKU·ACC-02Accounting & Invoicing

Accounting & Invoicing Planner

Research-backed planner that scopes bookkeeping and invoicing builds from 37 real accounting products — no scope bloat, no fabricated features.

37 sources · frequency-ranked
$29View planner →
SKU·JOB-03Job Board

Job Board Planner

Turn 'build me a job board' into a scoped, evidence-backed CLAUDE.md spec — without ATS bloat or resume-upload security gaps.

12 sources · frequency-ranked
$29View planner →
SKU·APP-04Appointment Booking

Appointment Booking Planner

Research-backed planner that turns a vague booking-app idea into a scoped, agent-executable spec across 7 service subtypes.

$29View planner →
SKU·INV-05Inventory Management

Inventory Management Planner

Turns vague stock-control prompts into a scoped, audit-ready build spec — without POS, ecommerce, or WMS scope bleed.

$29View planner →
SKU·RES-06Restaurant Management

Restaurant Management Planner

Structured discovery planner that turns a vague restaurant POS prompt into a scoped, evidence-cited CLAUDE.md or AGENTS.md build spec.

59 sources · frequency-ranked
$29View planner →
SKU·LEA-07Learning Management

Learning Management Planner

Research-backed LMS planner grounded in 38 real platforms — forces the right questions before your agent writes a line.

38 sources · frequency-ranked
$39View planner →
SKU·ONL-08Online Store

Online Store Planner

Research-backed ecommerce planner that turns 90-source feature counts into an agent-executable one-store spec with hard checkout and inventory gates.

90 sources · frequency-ranked
$29View planner →
SKU·HEL-09Helpdesk Ticketing

Helpdesk Ticketing Planner

Turn 'build me a helpdesk' into an agent-executable spec for a self-hosted support desk that actually replaces a paid subscription.

18 sources · frequency-ranked
$39View planner →
SKU·SCH-10School Management

School Management Planner

Turn a weak school-system prompt into an agent-ready blueprint grounded in 46 real school ERP and SIS products.

46 sources · frequency-ranked
$39View planner →
The method · §03

One file. Three moves. A spec your agent can actually ship.

Point your agent at the planner

Drop the planner into your repo and tell Claude Code, Cursor, or Codex to read it. No new tools, no lock-in — it's a CLAUDE.md / AGENTS.md your agent already understands.

add to repo

It interrogates you with cited checklists

The planner forces the decisions builders skip — billing model, roles, edge cases — and shows each feature with its evidence: “in 22/31 tabulated products.” You answer; it records the rationale.

forced questions

You get a complete, executable spec

Out comes a full specification your agent runs end-to-end — scoped, sequenced, and traceable. No mid-build hallucinations, no “demos fine, dies in prod.”

ship the build
Why it holds up · §04
80–90%

of AI agent projects fail to reach production. We start from what shipped — not from what a prompt imagined.

Source — RAND, 2025

50+

real, in-market products closely reviewed per vertical — drawn from hundreds surveyed across OSS starters, live SaaS, and marketplace templates, by hand.

Per planner research

Every·row

cited. Each checklist item traces to an inventory entry — the one thing your own AI can't fake without redoing the scrape.

Curation guarantee

$29–$49 per planner. Lifetime access, free updates as the inventory grows. No subscription, no lock-in.

See all pricing
Forge with evidence

Buy the inventory you'd never have time to build.

Stop shipping demos that die in production. Start from a spec grounded in what real products actually do.