Deal One-Pager
The executive summary, verdict and headline numbers, ready to forward.
Drop a deck or data room. About a minute later you have a verdict, investment memo, term sheet, IC deck and live model, with every headline figure traced to its page, its formula or a labeled assumption.
145 asset classesNo forms required
The problem
A Series A, a buyout, a private placement or a loan: someone still reads the deck and the data room, spreads the numbers, builds the model, stress tests it and writes the memo. That takes days per deal, so most deals get a skim, and the hard ones get skipped or rubber stamped.
AI underwriting software reads a deal's documents, extracts the financials, runs the analysis an underwriter would (cash flow or unit economics, valuation, structure, returns and risk) and drafts the deliverables: an investment memo, a term sheet and a model. The best tools trace every number to its source and label anything they had to assume, so a person can check the work.
Underridge is AI underwriting software for anyone who underwrites: investors, lenders, acquirers and advisors. Venture rounds, buyouts, private placements and structured deals, real estate, private credit, royalties, long-lived assets, commodities and insurance-linked deals: 145 asset classes in all. It is decision support: a person still owns the decision.
How it works
A pitch deck, a zipped data room, a model or a forwarded email. No intake form, no template, no setup.
Underridge loads the playbook for that asset class, reads every page and tests the sponsor's claims. Each headline figure is tied to its page, its formula or a labeled assumption.
Proceed, proceed with conditions, or pass, with the reasons. Then seven files: one-pager, term sheet, investment memo, IC pack, timeline, sensitivities and model.
What you get
One run produces the whole package a committee expects, built from the same evidence so the memo, the terms and the model agree.
The executive summary, verdict and headline numbers, ready to forward.
LOI-style indicative terms sized to the underwriting, with the binding clauses.
Thesis, cash flow or unit economics, valuation, structure, risks and conditions.
The investment committee presentation: thesis, numbers, risks and conditions.
A 90-day path to close: gates, critical path and diligence costs.
Two-way tables across the variables that actually move the answer.
A live Excel model with linked formulas. Change an input and IRR, MOIC and coverage recalculate.
Sample
A sample real estate loan, exactly as Underridge writes it. Same engine for equity, venture and structured deals.
Every headline figure carries its source, for example "T12 p.3", or is labeled as derived or as an assumption. Anything missing from the data room is flagged, not invented.
Proof
real deals taken from documents to a finished package.
median from data room to the full package.
of production runs labeled document-grounded.
deliverables generated: memos, term sheets, IC packs and models.
asset classes, on 7 model archetypes, from venture rounds and buyouts to real estate, royalties and loan pools.
Production figures across 174 underwrites; verdict mix across all 178 runs. Underridge says no more often than yes.
Agent-native
The same engine as the app, exposed three ways. One credit per underwrite, results in about a minute.
Point Claude or any MCP client at the Underridge server. Tools to underwrite, poll status and read the full package.
Queue a deal with its documents, poll the id, then fetch the memo, terms and model inputs as JSON.
Finds decks, data rooms and CIMs in your Gmail, adds them to the pipeline and saves tailored replies as drafts. It never sends mail.
// mcp.json: connect any MCP client
{
"mcpServers": {
"underridge": {
"url": "https://app.underridge.com/mcp",
"headers": { "Authorization": "Bearer ur_live_..." }
}
}
}
// then, from your agent
underwrite_deal({
deal_name: "Series A, $14.0M round",
files: [{ name: "Pitch_Deck.pdf", url: "https://..." }]
})
// queued, about a minute, then get_deal for the verdict and cited figures
# queue an underwrite (one credit)
curl https://app.underridge.com/v1/underwrites \
-H "Authorization: Bearer $UNDERRIDGE_KEY" \
-H "Content-Type: application/json" \
-d '{
"deal_name": "Industrial services buyout",
"files": [{ "name": "CIM.pdf",
"url": "https://..." }]
}'
202 { "id": "...", "status": "queued" }
# poll, then read the package
GET /v1/underwrites/{id} # queued, running, done
GET /v1/deals/{id} # memo, terms, model, sources
# connect Gmail once, then sweep the last 90 days
sweep newer_than:90d (decks, data rooms, CIMs, share links)
found deal emails with attachments or data room links
match each one to your Underridge pipeline
add new deals as pipeline entries
draft a tailored reply in Gmail, saved as a draft
# never sends mail · no credits spent
# batch underwrite the pipeline when you are ready
Pricing
Credits from $9.80 down to $5.95 per underwrite. Committed volume from $5.00. Every plan covers every asset class.
Your data room is encrypted in transit and at rest.
FAQ
AI underwriting software reads a deal's documents, extracts the financials, runs the analysis an underwriter would (cash flow or unit economics, valuation, structure, returns and risk) and drafts the deliverables, such as an investment memo, term sheet and model. Good tools cite the page behind every figure and label anything they had to assume, so a person can verify the work.
A verdict of proceed, proceed with conditions, or pass, with the reasons, followed by seven files: a deal one-pager, term sheet, investment memo, IC decision pack, deal timeline, sensitivity tables and a live Excel model workbook.
The verdict usually lands in about a minute. Across 174 real deals, the median time from data room to the full seven-file package is 76 seconds.
From your documents. Every headline figure is cited to its source page, for example T12 p.3, or labeled as derived or as an assumption. Anything missing is flagged, not invented.
No. Across 178 production runs, 56% of verdicts were pass and 44% were proceed with conditions. There have been zero unconditional approvals.
Any capital decision: venture and growth rounds, buyouts from the lower middle market up, private placements and structured deals, real estate, private credit and loans, royalties, long-lived assets, commodities and insurance-linked deals. That is 145 asset classes, built on 7 model archetypes.
Yes. Underridge exposes an MCP server with an underwrite_deal tool, a REST API (POST /v1/underwrites) and an inbox sweep that turns deal email into a pipeline. Each underwrite uses one credit and runs on the same engine as the app.
New accounts get 10 free underwrites. After that, credit packs run from $9.80 down to $5.95 per underwrite: Starter is $49 for 5, Professional is $179 for 25 and Desk is $595 for 100. Enterprise committed volume starts at $5.00 per underwrite.
No. It does the reading, spreading, modeling and first drafts, which is where most of the hours go. A person still owns the decision, the negotiation and accountability for the call. Underridge is decision-support software.
Get started
Drop it in. The verdict lands in about a minute, with the full package right behind it.
Underridge is a decision-support tool. Its output is not an investment or credit decision, an appraisal, or investment, legal, or tax advice, and it does not replace independent professional judgment or required diligence.