Capital Refinery
For private credit

Know how long until a covenant breach matters.

Capital Refinery parses the credit agreement, binds the covenants to the live operator data, and runs continuous breach probability and time-to-consequence on every position. The diagnostic the credit monitoring tools were never built to be.

The credit workflow today

Five gaps direct lenders are operating around.

Bain estimates private credit AUM at $1.7 trillion and growing. The infrastructure for governing those positions is still spreadsheets and PDF compliance certificates — and, per the EY 2024 Private Capital Tech Survey, the dominant blocker is the absence of a unified data architecture, not model capability.

What teams think they have
What actually happens
01A covenant model that updates with each operator submission.
A spreadsheet copy of the covenant table maintained by hand and three months out of date.
02Continuous breach probability across the book.
A quarterly compliance certificate that arrives after the breach has already happened.
03Time-to-consequence ranking on every position.
A status colour on a dashboard that does not know how much room is left.
04Stress shocks that propagate live across the portfolio.
An ad-hoc Excel rebuild every time the LP asks for an interest-rate scenario.
05Lender consent rights surfaced as workflow, not legalese.
A 200-page agreement nobody on the monitoring team has actually read since closing.

Bain & Company 2025 Global Private Equity Report; EY 2024 Private Capital Tech Survey

Portfolio-wide stress event
Per position, every time
Trigger
Rates move 50bps · Sector EBITDA shock · LP request
01
Pull credit agreement
Find the PDF in the deal folder
02
Rebuild covenant model
Re-spread financials in Excel
03
Re-run the shock
Apply rate or EBITDA delta by hand
04
Hand-format for IC
Write the memo, build the slide
Loop ↺
Repeat for the next position in the book. Then the next. Then the next.
A rate move or EBITDA shock forces the credit team to manually reconstruct the agreement context on every position in the book — sequentially.
MANUAL · 0% reuse
What the platform does for credit

Three engines built specifically for direct lending.

01 / 03

Credit agreement parser

Three-pass deterministic + LLM + human review.

Facilities, tranches, lenders, covenants, definitions, and consent rights are parsed into a structured capital structure record. Every clause is mapped to a workflow surface and traceable back to the agreement page.

02 / 03

Covenant engine + PIK + waterfall

The covenants are alive, not a tab in a spreadsheet.

Each covenant is bound to the operator data that tests it. Breach probability runs continuously. PIK accrual, payment waterfall, and lender consent are first-class workflows — not memo footnotes.

03 / 03

Time-to-consequence ranking

Sort the book by how much time is left.

Every position gets a live time-to-consequence score. The team sees, at portfolio level, which positions are operating closest to a structural break — and how long they have to act before the window closes.

Bring us a credit agreement.

We'll parse it into a structured capital structure record and run breach probability against the live operator data. Real document. Real position. 60–90 minutes.