Sheet U-101 · Underwrite
Pro forma pro_forma
Check an existing pro forma or model a new deal. Your agent rebuilds the deal in the engine and reports where its numbers differ from yours.
Runs from your AI agent: Claude Code, Codex or Claude Desktop.
10 to 25 credits ($0.10 to $0.25) per run
What it does
| Model verification | An existing pro forma rebuilt in the engine, with each metric that differs from yours reported as a finding |
|---|---|
| Built-in checks | Flags modelling errors, such as a double-counted vacancy or a loan balance left open at sale, each with a code and the input it refers to |
| Repeatable results | The same input gives the same numbers and the same run id |
| Capital stacks | Construction loans that refinance, partners in funding order, and waterfalls with pref, catch-up, hurdles and promote |
| Excel workbook | Presentation, model and provenance sheets, saved to your computer |
Two ways to use it
A · Most common · Check an existing pro forma
Hand your agent the Excel workbook, or a PDF of it. Your agent rebuilds the deal in the engine and lists the engine's numbers beside yours; each difference is a finding.
B · Model a new deal
Describe the deal in plain words, or point at a term sheet. Your agent writes it out and the engine models it.
Model a 48-unit garden apartment purchase with a 65% loan at 6.25%, a 5-year hold and an 8% pref with a 70/30 split above it. Show me levered IRR for the LP and the GP.
What it takes in
One deal description as JSON. Your agent builds it from your pro forma or your description and checks it against the published schema first.
| Architectural program | Rental multifamily, for-sale condo, retail, office, industrial, hotel, short-term rental, parking and land, alone or mixed |
|---|---|
| Timeline | Fixed dates, or dates set from events such as closing, delivery, stabilization and sale |
| Budget assumptions | Land, hard, soft, financing and fee costs, each with its timing |
| Capital stack | Debt with sizing, rate, interest-only, amortization and refinance; equity held by named partners, in funding order |
| Distributions | Operating and sale waterfalls with return of capital, catch-up, hurdle tiers and splits, or a standard preset |
A run stops at 30 seconds; typical deals finish in well under a second.
What you get back
Metrics for each perspective you ask for: levered and unlevered IRR, equity multiple, cash-on-cash, yield on cost, development spread, stabilized NOI and cap rate, DSCR, debt yield, loan to cost and to value, peak equity and residual land value. Diagnostics name the input each finding refers to, so your agent can fix it and run again. A run id identifies the run.
Price
| Scheme | Input complexity |
|---|---|
| What sets the price | 10 credits, plus 2 for each product after the first and 2 for each waterfall tier after the first, up to 25 |
10 to 25 credits ($0.10 to $0.25) per run. A run that fails because of its input or script costs 1 credit; nothing is charged when the failure is ours. See pricing for credit packs.
The price is worked out before the run. Waterfall tiers are counted in the largest waterfall.
| Deal | Credits | Price |
|---|---|---|
| 1 product, 1 waterfall tier | 10 | $0.10 |
| 2 products, 1 waterfall tier | 12 | $0.12 |
| 3 products, 2 waterfall tiers | 16 | $0.16 |
| 5 products, 3 waterfall tiers | 22 | $0.22 |
| 8 or more products with several waterfall tiers | 25 | $0.25 |