consultance.ai
Finance and data

Private Credit Underwriting Pack

A direct-lending credit desk that runs inside Claude on your own term sheets and data rooms, for the price of a Claude seat instead of a $300K analyst and a six-figure third-party diligence engagement. Read the data room, build the cash flow and DSCR model, stress test covenants across rate paths, flag PIK creep, score downside and recovery, and draft the IC memo, every figure tied to a source. Direct lending diligence that used to run 3 to 6 weeks now runs in two days, with a source-tie gate so nothing untied reaches committee.

A $300K analyst grinding the data room, or two days in your own Claude.

Direct-lending diligence that runs 3 to 6 weeks now runs in two days. This is a credit desk inside Claude on your own term sheets and data rooms — DSCR model, covenant stress test, PIK-creep flag, recovery score, and the IC memo, every figure tied to a source — for a Claude seat instead of a $300K analyst and a six-figure diligence engagement.

Built on open source
OpenBB (open-source data layer)

The code is public and free. The setup prompt below installs and wires it for you.

Step 1

Paste this setup prompt — Claude installs it for you

Easy mode · paste this into Claude

Claude installs it for you, step by step.

Never used Claude before? It is free to start. Open it in a new tab, copy the prompt, paste it in. It asks one question, then walks you through everything.

  1. Step 1
    Open claude.ai ↗

    Sign up free. No card. Takes 30 seconds.

  2. Step 2

    One click. Lands on your clipboard.

  3. Step 3
    Paste + send

    Claude asks what you need + guides you the rest of the way.

Open claude.ai ↗
Tune the prompt for your level (optional)
Preview the prompt (you do not need to read it)
I want to run private credit underwriting in Claude on my own deals, replacing the data-room grind and the third-party diligence engagement. Walk me through it, step by step.

This is NOT a Terminal or coding install. The prompts run inside the Claude app. Treat me like a credit analyst, underwriter, or fund principal who has never coded. The only optional technical bit is connecting a free public-data source, and I will give you a no-install fallback for that too.

What I am setting up: a private Claude Project that holds the deal data room, so the vault on this page can read the term sheet, build the model, stress the covenants, flag the PIK creep, and draft the IC memo, all on my own deal. Nothing installs on my machine. A named human signs the credit decision before committee.

Walk me through ONE step at a time. Wait for me to confirm each before moving on.

1. **Plan + privacy.** Open `claude.ai`, sign in. For real deal data use a Team or Enterprise plan so the data room stays in my own tenant and is excluded from training. Confirm I am on one before uploading anything confidential. Pin Opus 4.8.

2. **Create the deal workspace.** Left sidebar → "Projects" → "Create Project". Name it for the deal or my book (for example "Project Atlas Credit" or "Credit Desk Q3").

3. **Load the data room into Project Knowledge.** Upload the term sheet, the financials, the model, the credit agreement, existing debt schedules, and any prior filings. This is the data source option (A) in prompt 01. For a quick one-off, I can paste figures instead (option B).

4. **Optional, public comps and market data.** If I need public comparables or rate data, OpenBB (`github.com/OpenBB-finance/OpenBB`) is a free open-source data platform a dev on my team can run, or I paste pulls in. If my fund licenses a market terminal, connect it via the Claude for Financial Services connectors. This is option (C). Skip it if my deal is self-contained in the data room.

5. **Set Custom Instructions.** In the Project's Custom Instructions box, paste: "You are a direct-lending credit desk supervised by a named human. Every figure on a memo ties to a data-room source or a pulled figure. Never invent a number, covenant level, or recovery assumption. Flag PIK creep and untied figures. A human signs the credit decision before committee."

6. **Run prompt 01** (the router) from the vault on this page. Answer deal type, where the data lives, and the context. Then run 02 to 10 in order: read the room, build the DSCR model, stress covenants, flag signals, score recovery, reconcile the ledger, summarize the agreement, draft the IC memo, and run the source-tie auditor.

7. **Confirm the output bar.** Check that every figure ties to a source, PIK and covenant risks are flagged, and prompt 10 blocks any untied number before the memo ships.

## Rules for walking me through this

- One step at a time. Tell me exactly what to click or upload.
- Define jargon once: Project, Project Knowledge, Custom Instructions, data room, DSCR, FCCR, covenant headroom, PIK, LGD, recovery waterfall, `<review_gate>`, `{{TOKEN}}`.
- Never tell me to use a Terminal for the prompts. The whole vault runs from a browser or the Claude desktop app.
- Anti-pattern call-outs: if my firm blocks claude.ai, use Claude Enterprise with SSO; if I cannot find Projects, paste prompts into a normal chat and paste the data room in pieces; if the data room is huge, tell me to upload the term sheet, financials, and credit agreement first and add the rest as the prompts need it. Do NOT tell me a step is "not possible", give me the upload-in-pieces or paste fallback.
- A named human signs the credit decision. Claude builds the model and the memo and ties every source; a person signs before committee.

First message: ask me "What deal type, and where does the data room live, uploaded into a Project, pasted, or pulled from a connector? And are you on a Team or Enterprise plan so the deal data stays in your tenant?" Then start step 1.

When the data room is loaded AND prompts 01 to 04 run on a real deal (read, model, stress), switch into "desk mode" and help me wire my standard covenant set and IC memo format so every future deal runs the same way.
Step 2

Step 1 installed it. Now run these 10 prompts on your own data.

the vault

The 10 prompts

Tap a prompt to jump to it. Hit copy. Replace the tokens. Paste into Claude Opus 4.7.

where it breaks

Before you connect live data

  • • Run dummy data first. Real client data is not a test bed.
  • • API keys never go in a public repo. Use env vars and a secrets manager.
  • • Add logging, access control, monitoring, and a rollback path before launch.
  • • Read the license. Forking a repo without checking is how lawsuits start.
license note

Credit the original author

Prompt set authored by consultance.ai. Apollo, Ares, Blackstone, KKR, Oaktree, and Carlyle are referenced as the standard the desk matches, no affiliation implied. OpenBB is a free open-source data source used under its own license. Your deal data stays in your own Claude tenant; we never see it. This is not investment or credit advice; a named human signs every credit decision before committee.

the newsletter

AI news worth opening.

The AI tools, launches, and shifts that actually matter, in plain English. New library drops the moment they land.

100% freeNo paywall, everUnsubscribe anytime

Read this far? You want Apollo-grade diligence in two days, not six weeks. Let us build the desk — source-tied and signed before committee.