Latest in the build log: Founder page carries its approved final text

product

DealDesk Underwriting Desk

One desk for a real-estate deal, from first look and underwriting to the investment committee and the post-purchase review. Residential deals are read through mortgage payment and cash-on-cash; commercial deals through IRR, DSCR and NOI. Every key number records where it came from and who checked it. Missing numbers show as "—", never as 0. MIT-licensed and self-hosted: your data stays on your own server.

features

What it does

Three tracks: residential, commercial, institutional

Residential deals are read through payment and cash-on-cash; commercial deals run from rent roll and T12 to NOI, cap rate, DSCR and IRR; institutional deals are modeled monthly, lease by lease, with probability-weighted rollover, downtime and free rent, TI / LC, NNN or base-year recoveries, floating debt with a rate cap and an LP / GP waterfall, and export a formula-linked Excel model and an institutional IC memo. Every run has an ID anyone can recompute.

A passport for every number

Key numbers record the source file, page, confidence and reviewer. With sign-in on, whoever entered a number cannot also verify it (four-eyes rule). Figures that disagree across sources are listed as conflicts and not trusted until resolved.

Verify first

Other tools tell you how many items are unverified. DealDesk ranks each one by how many basis points it can swing the main return metric, so the first line is the document to chase first.

A devil's-advocate memo, without AI

Before committee, the system writes how this deal is most likely to fail: kill criteria it barely clears, conclusions resting on one assumption, a losing downside scenario, unverified numbers. Each point comes from fixed rules, is reproducible, and is left out when the data is missing.

Max bid, solved backwards

All kill criteria (DSCR floor, cash-on-cash, IRR and more) are applied at once to solve for the highest price you can still pay, and the binding constraint is named. Drag the price to see how much room each floor has left.

From committee to post-mortem

Committee packs are generated in one step, with a chairman veto. After purchase, enter actual NOI and see which underwriting assumption the variance comes from. Roles are chairman, analyst and viewer; every change lands in an audit timeline.

deploy

How to run it

The live app is the founder's own production instance (an N=1 proof). To self-host: Python 3.11 or later, then ./run.sh starts it locally; use Postgres in production. The interface switches between Chinese and English, and the 68 REST endpoints come with an OpenAPI spec.

Project status

Live & ongoing

Latest milestoneInstitutional underwriting is live in production: lease-by-lease and unit-mix (multifamily) monthly cash flows, floating-rate debt, an annual or monthly LP / GP waterfall (catch-up, clawback), and exports of a formula-linked Excel model and an institutional IC memo 2026-10-24

Next stepShip the next batch: development projects and construction loans, mezzanine and refinancing, Google sign-in (card follows the measured production reading after deploy)

The live app is the founder's own production environment, also open to invited friends: invitation only, no public sign-up, no charge. This page only displays its status. No return or performance claims.

Data as of 2026-10-24

Next

To see how it is being built, read the changelog; or go back to the business overview.