What gets built — on the stack we already run.

Next.js (App Router) + TypeScript + Tailwind, Supabase (Postgres, row-level security as the tenant boundary), Render (web service + background workers), Square (billing), Resend (email). The same stack behind Funding Depot, We Know Restaurants, We Know Salons and Tie Out Bureau — one operator can run all of them because none of them ask for anything the others don't already use.

The company's build — two months, $50,000, per the business plan §08, to get the templated engine, CRM, card and dashboard built once. Separate from that, each agent's own turnaround splits into two pieces that used to get collapsed into one "live in a week" promise, and shouldn't be:

The second number isn't a KeyRingOS estimate — it's Placester's own published turnaround for the same IDX approval process, and KeyRingOS goes through the same boards. Worth designing around rather than promising away: the site, the card, the bio and the lead-capture form don't need to wait on the MLS at all, and shouldn't. The agent's card and site go live first, fully functional for lead capture and contact, with listings populating as soon as the board approves — not a single launch gated on the slowest part of the process.

Placester is the nearest real competitor (§03 of the business plan) and the clearest published feature set to build against. Its CRM: autoresponders, email blasts, drip campaigns, lead capture forms and pop-ups, activity tracking, lead scoring, all built around a drag-and-drop website builder. KeyRingOS matches the CRM feature-for-feature and drops the builder — there's nothing to drag, because the site is templated and MLS-connected already — and adds two pieces nothing in §03 of the business plan bundles together with a CRM at all: the card, and a real transaction document/e-signature layer.

Worth separating clearly, because they don't work the same way. The agent's own listings — a handful of active properties — are what the admin panel actually curates: pulled from the MLS feed, shown on the site by default the moment they sync, with a per-listing hide switch and bulk show/hide for anyone who wants everything off until reviewed. Automatically archived the moment the board marks one sold, withdrawn or expired — never a stale "active" listing sitting on a live site. The broader IDX board search — every other property on that MLS, via reciprocity — is the second surface, and nobody curates that one listing by one; it's what powers site-wide search for buyer leads, generated automatically, same as every IDX competitor in §01.

Real Geeks' EstateIQ and Zillow's Zestimate both already do the instant, computer-generated estimate — free, ubiquitous, and the reason competing on speed or accuracy against them is pointless. KeyRingOS's version doesn't try: no AVM vendor, no per-lookup cost, no data integration beyond what's already here. The pitch flips instead — "see what your home is actually worth," positioned against the generic algorithmic number, delivered by an agent who looks at the real property, not just public records. The form itself carries that promise: address, plus the details an algorithm can't see — condition, recent renovations or upgrades, bedroom/bathroom count if it's changed since the public record, and timeline to sell. That's the same "Basic lead capture" mechanism as the tick above, just asking more — a longer form than a Zestimate-style tool would ever need, which is exactly what makes the resulting lead both more qualified and more credible than a one-field address box. Lands as a seller lead, a CRM record like any other lead capture form; the agent prepares the actual number as a real CMA.

KeyRingOS doesn't author or generate any of these documents — real estate paperwork is state- and association-standard, and pretending to write legally binding forms would be both wrong and unnecessary. What an agent has (a listing agreement from their MLS, a disclosure from their state association, a purchase agreement drafted for a specific deal) gets uploaded, tagged against a 27-type taxonomy across five categories that match the real chronology of a transaction (onboarding & agency agreements, disclosures & advisory forms, offers & core contracts, contingency management & amendments, closing day & settlement), tied to the client it belongs to, and processed the way that specific document actually requires:

Two real techniques, not one black box. A vision-capable model reads the uploaded document and classifies which of the 27 types it is, then proposes where the signature fields likely belong from layout and text cues — never trusted blindly for something legally binding, so the agent confirms or adjusts placement before anything sends. Once confirmed, that exact form's layout is fingerprinted and saved. The next time any agent uploads that same standard association form — which happens constantly, since most agents in a state or brokerage use the same handful of templates — KeyRingOS recognizes it and places the fields automatically, no confirmation needed. It gets faster because real estate paperwork is repetitive, not because of an abstractly "self-improving" model.

Worth being precise here rather than rounding up: once a board is connected, Closing Boss's own cost per additional agent on that board is $0 — that's already the core of the margin story in the business plan (§05, §07). What isn't free is connecting to a new board for the first time — SimplyRETS charges $99 one-time per board, plus a subscription on Basic tier (the working default — see §07 of the business plan for why Basic and not Premium): $49/mo for the first board, $25/mo for each additional one, a tier chosen for feature coverage rather than a figure that creeps up with volume on its own. Real numbers, confirmed directly against the account, already priced into §08's cost tables — not invented for this document.

Not "give KeyRingOS API access" — an agent doesn't have API access to hand over. What has to be true: the agent already holds an active MLS membership (their own license and dues, $25–200/mo depending on the board), and separately, whatever IDX fee that board itself charges to participate in IDX at all — commonly $20–50/mo, sometimes a one-time IDX setup fee plus a smaller annual fee instead. Both are the agent's own pre-existing cost of being online at all, paid to the board directly, owed regardless of which platform they choose — not something Closing Boss bills, covers or has any part of. KeyRingOS then requests the technical credentials from the board, using that membership plus a signed IDX authorization as justification — KeyRingOS holds the resulting API credentials, not the agent. Same shape Placester uses, and the same boards, so the same approval timeline applies.

Worth being explicit, because it's easy to read "curated" as "the agent selects listings from the board by hand." It isn't that. SimplyRETS's API takes an agent parameter that filters results down to one listing agent's own inventory — "My listings" (§03) is that query, scoped automatically the moment the board connects. The agent's only manual input is the per-listing hide switch, an opt-out on top of an automatic feed, never an opt-in pick from the full board. The full board itself stays reachable through the separate, uncurated search surface — two different jobs, per §03's listing-admin split, not two versions of the same feature.

SimplyRETS is consumed as a JSON API, not embedded as a page — the distinction matters because most cheap IDX sites embed the MLS vendor's own canned search widget instead, and those are uniformly bad: dense multi-select dropdowns with hundreds of subdivisions, year-built spinners back to 1901, no map, no sliders, no saved search. Same underlying board data, a search UI KeyRingOS actually builds — map-first, live filters, autocomplete — because the data arrives as JSON to build against, not a widget to embed.

One plan, already priced in the business plan (§07) — restated here as product scope rather than re-derived:

Every agent site is built to be indexed properly from day one — real per-listing pages, not a single JavaScript-rendered blob — and SEO is one of the two channels the marketing budget actually funds (business plan §08), alongside paid search. Not a separate product or a service tier; every agent's site benefits regardless of which pricing tier they're on. The blog (ia.html §05) is where this actually shows up as content: KeyRingOS-syndicated market posts, displayed under each agent's own site, not something an agent writes themselves. Content marketing as a platform commitment, not a to-do list item added to every agent's plate.