The Private Credit Covenant Compliance Checklist

1. A Checklist Is a Discipline, Not a Form

Covenant compliance is where a lender's protection is either enforced or quietly lost. Not usually in one dramatic breach, but in the small item skipped on the twentieth loan during a busy week.

A credit agreement can be perfect on the page and useless in practice if nobody checks it on schedule. The covenant that hurts you is rarely the one you were watching. It is the reporting deadline that slipped, the add-back that crept into a certificate, the basket that filled while attention was elsewhere.

That is what a checklist is for. Not a form to file, but a promise about what will never be skipped, no matter how full the week. This guide is the list a private credit monitoring team actually works through, with the reasoning behind each item, so it holds up as the book grows. It sits inside the wider program in the complete guide to AI for private credit, and gets specific about the one thing examiners, LPs, and your own downside all care about: proving, every period, that each borrower is inside its covenants.

2. How to Run It: Extract Once, Check Every Period

Run the checklist in two phases, and keep them separate.

The first phase happens once, at onboarding. Read each credit agreement and extract its own terms into a structured record: the covenant levels, the definitions they are measured against, the baskets, the cure rights, the reporting deadlines. A credit person validates that record against the document. From then on you measure the borrower against the stored record, not against a fresh read of a two-hundred-page agreement every quarter.

The second phase repeats every reporting period. The borrower's financials arrive, you calculate each ratio, check the certificate, update headroom, and escalate anything that moved. Same steps, same order, every time.

The division of labor is identical in both phases, and it is worth stating once so the rest of the checklist can assume it. AI extracts the terms, calculates the ratios, and flags the exceptions. A credit person concludes and signs. The model gives you coverage across a larger book and a drafted position on every loan. The judgment stays human. The mechanics of standing this up, the workspace that holds the agreements and the recurring check, are in Claude for covenant monitoring.

3. Financial Covenants: The Numeric Tests

Financial covenants are the numeric tests, and they are the heart of the checklist because they are the ones measured every period.

There are usually a handful. A leverage ratio, net debt to EBITDA, tested against a maximum. An interest coverage ratio, EBITDA to interest, tested against a minimum. A fixed-charge coverage ratio, earnings against interest, scheduled principal, and rent. A debt-service coverage ratio in cash-flow and asset-based structures. Sometimes an absolute minimum EBITDA floor, common in lighter structures with no ratio cushion. And often a minimum liquidity level, a floor on cash or availability.

For each, check three things: the level, the definition it is measured against, and the test date. The level is easy. The definition is where the risk hides. The EBITDA a leverage covenant is tested against is the contractual one, built from a negotiated definition with add-backs and caps, and it differs from the loan next to it. Get the definition wrong and every ratio built on it is wrong.

Here is what a monitoring team watches for across the common financial covenants.

AI earns its place here. It extracts each covenant level and its definition once, calculates the ratio from the borrower's financials every period, and flags the ones drifting toward their limit. What it does not do is decide whether a soft quarter is noise or the first crack. That call, and the read on an aggressive add-back, stays with the credit professional. The depth on extracting these definitions is in AI for credit agreement and covenant review, and the productized version of the whole extraction lives in borrower intelligence.

4. Affirmative and Negative Covenants

Affirmative covenants are the promises to do something. Deliver financial statements on time, maintain insurance, pay taxes, keep the corporate existence intact, allow inspections, and give notice within a set window of a default, a material lawsuit, or a change of control. They are easy to check and easy to forget, which is exactly why they belong on a list.

Negative covenants are the promises not to do something, and they come with baskets and carve-outs. Limits on incurring additional debt, granting liens, making restricted payments like dividends and distributions, selling assets, making investments and acquisitions, and transacting with affiliates. The protection is real, but it leaks through the carve-outs. A generous basket lets the borrower do quietly what the covenant appears to forbid.

For each, check whether the borrower has stayed inside the basket, and track cumulative usage, because baskets are often measured as a running total or a builder that grows over time. The notice covenants need a trigger of their own: some obligations only start when an event happens, so the checklist has to watch for the event, not just the deadline.

AI extracts the baskets and their caps, tracks how much room is left, and flags when a basket is near full or an action looks off-covenant. Whether a given move is aggressive or fine, and whether to raise it with the borrower, is the credit team's call.

5. Reporting Obligations and Deadlines

Reporting obligations are the covenants that feed all the others. If the financials do not arrive, you cannot test anything, so the calendar is its own checklist item.

A typical loan requires monthly financials within about thirty days, quarterly financials and a compliance certificate within about forty-five, and annual audited statements within roughly ninety to one hundred twenty. Add the annual budget, and event-driven notices due within a set number of days of the event. The exact windows differ by loan, which is the point: each one goes in the calendar with its own deadline.

Check three things every period: did the reporting arrive, was it on time, and was it complete. That sounds trivial until you are tracking dozens of borrowers on different cycles.

Late reporting is itself an early warning. A borrower that always reported on the fifteenth and is now silent on the twenty-fifth is telling you something before the numbers do. AI extracts each loan's deadlines into a live calendar and flags what is late or missing, so the exception list starts before anyone opens a financial statement. What a given delay means for a given borrower is a human read. Setting this tracking up from scratch is covered in how to set up AI covenant monitoring.

6. The Compliance Certificate Review

Every period the borrower sends a compliance certificate: its own calculation showing it met the covenants. The lender's job is to check that certificate, not take it on faith.

Work through four questions. Does the EBITDA build use only the add-backs the agreement permits, or has an unapproved category slipped in? Does each ratio use the contractual definition, not a rounder version of it? Do the figures tie to the financial statements behind them? And is the certificate signed by the officer the agreement names?

This is the check that catches the quiet problem. A certificate can pass its own internal math and still be wrong against the agreement, because it used a definition that is close but not the one negotiated. Over a book, that is the difference between checking every certificate and hoping the ones you skipped were fine.

AI reconciles the certificate against the agreement and the statements, and flags the line that does not tie, with a pointer to the clause and the number it used. The reviewer confirms the flag in seconds instead of rebuilding the calculation. The sign-off is still the reviewer's, and no figure drives a decision until a person has traced it to its source.

7. Headroom Thresholds and Proximity Alerts

A pass or fail tells you the least useful thing: where the borrower was on one date. Headroom and its direction tell you where the borrower is going.

For each financial covenant, calculate the remaining headroom, how far the borrower sits from the level, and track which way it has moved over the last few periods. Then set the alert bands. A firm might treat anything under, say, fifteen percent headroom as amber and a breach or near-breach as red. The exact bands are a choice; having explicit ones is not.

Headroom is a direction, not a number. A borrower at 4.1x against a 4.5x covenant, slipping a tenth every quarter, is a conversation you want two quarters early, not the morning the certificate lands. A borrower steady at 4.3x for two years is calmer than the raw number looks.

AI calculates the headroom, ranks the book by proximity and by trend, and surfaces the names tracking toward a level in the next couple of periods. That turns a manual sweep of every loan into a short list that needs a human to look closely. Reading the flag, and deciding whether the trend is structural or seasonal, is the analyst's job. The portfolio-wide version, and how it feeds risk rating and the watchlist, is in AI for private credit portfolio monitoring.

8. Cure Periods and Grace Mechanics

A breach is not one state. It is three, and treating them as one is how a covenant program trains its own team to ignore alerts.

The three states: compliant, in the cure window, and breached after the cure has lapsed. Reporting covenants usually carry a notice-and-cure grace period, so a missed deadline is not immediately a default. Financial covenants often carry equity cure rights: the sponsor can inject equity, counted toward EBITDA or against debt, to cure a breach. Those rights are limited, typically capped in number and blocked in consecutive quarters, so tracking how many remain matters as much as the breach itself.

For each exception, check which state it is in and, for equity cures, how many are left. A breach the day the statement arrives is not a breach past the cure window, and flagging all three the same way is the fastest route to alert fatigue.

AI extracts the cure terms per loan, tracks cure usage against the caps, and labels which state a given exception sits in. Whether to accept a cure, waive a technical breach, or enforce is the credit officer's decision, and it carries consequences for the mark, the LP report, and the relationship.

9. The Escalation Path When a Covenant Tightens

A flag that goes nowhere is worse than no flag, because it teaches the team that flags do not matter. Every amber name needs a next step and an owner.

The escalation path is a sequence, run in order. An analyst reviews the flag and confirms it is real. A confirmed concern goes to the watchlist and, past a threshold, to the credit committee. Someone owns the borrower conversation. Depending on what that surfaces, the firm reserves its rights, negotiates a waiver or an amendment, or moves toward enforcement and a workout. Each step has a named owner and a trigger, so nothing waits in a queue while a covenant tightens.

Check, every period, that each amber and red name has a next step assigned to a person, not sitting in a report. The escalation is only as good as its ownership.

AI drafts the exception summary and the position pack the committee reviews: the headroom, the trend, the certificate reconciliation, the history. It does not decide anything. Every step from the borrower conversation onward is human, and the accountability sits with the credit team. The service that builds this monthly and quarterly read into a firm's operation is portfolio risk monitoring for private credit funds.

10. The Two Lines: Data and Judgment

Two lines keep the checklist trustworthy. Cross either and the discipline is worth less than the paper it is written on.

The first is the data line. Credit agreements, borrower financials, and covenant schedules are confidential. They belong on a firm-managed account, not a personal login. Claude Team and Enterprise, the business plans, do not use your business data to train public models; the consumer tiers can use your chats unless you opt out.

Be precise about what that buys you, because the honest version holds up better under scrutiny. Team and Enterprise keep your data out of public model training. They do not make your chats disappear. Standard retention still applies to the chat product, and the zero-retention arrangements people cite are an API and Claude Code matter, not the chat window an analyst types into. So the rule is short: keep borrower data on Team or Enterprise, and write it into the firm's AI policy before the first agreement goes in.

The second is the judgment line. AI extracts, calculates, and flags. A credit officer concludes and signs, and every figure reconciles to a source before it drives a decision. Where these tools fit against the rest of the credit stack is in the best AI tools for private credit.

11. Where to Start

Do not roll this out across the whole book at once. Pick one book and ten loans.

Extract the terms from those ten agreements and have a credit person validate the extraction against the documents. That validation is the point: it is how you learn where the model reads well and where it needs a human, on your own agreements, before you trust it at scale. Then run one period end to end on those names: reporting logged, ratios calculated, certificates reconciled, headroom flagged, exceptions escalated. Compare the result to what your team produced by hand.

When the drafts match on the standard names and the flags on the odd ones are real, widen. A covenant program built this way is slower to stand up and far harder to break, because every item earns its place before the book depends on it.

If you would rather build it with someone than assemble it alone, an AI Readiness Sprint is where it starts: the data rule, the first set of agreements extracted and validated, and a covenant read you can run every period. From there we run it with you toward portfolio risk monitoring that watches the whole loan book without adding headcount.