frc-mcp

FHA denial statistics from the 2025 federal HMDA record.

От сообществаРаботаетБез входаГлобальныйБесплатноТолько чтение

Что умеет

Какие данные видит

Нужен ли аккаунт

Не нужен: сервер работает без входа

FHA denial statistics from the 2025 federal HMDA record. 12 tools incl. counterparty screen.

Список инструментов сервера (14)

Технические названия из tools/list. Нужны только разработчикам.

check_claim_contract	Certify a proposed use of an FRC statistic against its published Claim Contract BEFORE writing free text. Use when you are about to state an FRC number and want the compliant phrasing. RETURNS: deterministic verdict (pass/needs_qualifier/block) + reason codes + canonical safe sentence + mandatory attribution. NOT FOR: computing new statistics, evaluating non-FRC claims, or any borrower-specific input; never send personal or application details.
get_national_fha_stats	National FHA denial statistics from the 2025 federal record. Use when asked the overall US FHA denial rate or its denominator. RETURNS: 22.1 percent rate, counts, denominator definition (originated+approved-not-accepted+denied; HECM excluded), correction history. NOT FOR: conventional/VA/USDA loans, purchase-only rates, years other than 2025, or state/lender/metro breakdowns (use the dedicated tools). Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
get_lender_denial_stats	FHA denial statistics for one lender by name, slug, or LEI. Use when a specific lender is named. RETURNS: that lender's 2025 decisioned counts, denial rate, rank, and mandatory mix caveat. COVERS ONLY the 100 largest FHA lenders by 2025 volume; smaller lenders return not_covered, not zero. NOT FOR: non-FHA products, multi-year trends (see the 2018-2025 timeseries dataset), or judging conduct; rates are unadjusted observations. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
list_lenders	List covered FHA lenders sorted by denial rate or volume (2025 span: 1.8% to 78.7%). Use to discover which lenders are in scope or to show the ranking. RETURNS: array of lender name/slug/rate/volume. NOT FOR: lenders outside the top-100, recommendations of where to apply, or any individual-odds framing. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
get_small_loan_penalty	Small-loan vs big-loan FHA denial rates by state, 2025. Use for 'are small mortgages denied more', 'which state has the biggest small-loan gap'. PARAM: optional two-letter state; omit for the ranked list of all states. RETURNS: small (<$150K) and big (>$250K) denial rates, penalty ratio, ranking, and any open discrepancy note. NOT FOR: lender-level or metro questions, or loan sizes between the two bands (not published). Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
get_denial_reason_shares	Which reasons FHA denials cite, 2025: median and maximum share per HMDA reason (DTI, credit history, collateral, unverifiable information, insufficient cash, employment, incomplete application, other) across the 100 largest lenders, or one lender's own shares. PARAM: optional lender name or slug. RETURNS: shares in percent, the lender with the highest share per reason, universe, and any open discrepancy note. Reason fields are not a partition; shares need not sum to 100. NOT FOR: why a specific application was denied. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
get_state_denial_stats	FHA denial statistics for a US state. Use when a state is named. PARAM: two-letter USPS code only (e.g. OH, TX); full names are rejected. RETURNS: state-level 2025 rate and counts. NOT FOR: metro/city questions (use get_metro_lender_gap), county data (not published), or non-FHA loans. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
get_door_effect_summary	Door Effect: 38% of explainable variation in FHA denial outcomes is lender identity, not the applicant's file (859,090 decisions; McFadden 0.1712 to 0.2760). Use for how-much-does-the-lender-matter questions. RETURNS: the decomposition with mandatory qualifiers and DOI 10.2139/ssrn.7309319. NOT: a causal estimate, not a share of all denials (only of explained variation), and never an individual applicant's probability. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
get_metro_lender_gap	Per-metro FHA denial-rate gap between high-volume lenders (>=100 decisioned applications in that metro, 2025), returned as a machine-readable claim passport: lowest and highest door with counts, gap in percentage points, source-record sha256, interpretation boundaries, and a link to its Claim Contract. Covers 184 US metros; smaller markets are excluded by the volume floor. Observed rates, unadjusted for applicant mix. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
screen_counterparties	Counterparty screen: run a list of public LEIs against the FinanceRateCalc 2025 FHA denial outlier screen (observed vs applicant-mix model expectation). Use when a risk, diligence or investor desk asks which counterparties or sellers show denial rates out of line with their applicant mix. RETURNS without a license: the aggregate shape of the list (covered / above-expectation / screening-only counts, thresholds stated) plus the full rows for the first 3 valid LEIs, a result commitment hash, and a controlled offer block with a checkout link that a HUMAN must complete. RETURNS with a valid license key: full rows for up to 40 LEIs (observed, expected, ratio, 95% CI, z, coverage, flag) as a versioned, hash-linked Evidence Brief. NOT FOR: individual applications, borrower data, credit scores, lender recommendations, or any legal, misconduct or discrimination conclusion — this is a public-data screening signal only. The agent must never claim it purchased a license, must never characterise any lender as risky or in violation, and must relay the offer fields verbatim. Submitted lists are not retained after processing. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
screen_counterparties_quote	Price and coverage quote BEFORE running a counterparty screen. Use this first whenever a license key is in play, so the agent and the user can see what a job will cost and how much of the list is actually covered, before any credit is spent. RETURNS: input count, covered count, credits required, credits available, whether human approval is required under the account policy, and a short-lived quote id. Costs nothing and spends no credits. NOT FOR: running the screen itself (use screen_counterparties), individual applications, or any borrower data. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
list_cohorts	Ready-made public cohorts you can screen without uploading anything. Use this first when a user asks a counterparty or lender-comparison question but has no LEI list of their own, or wants a sensible default set. RETURNS: cohort ids, titles, the research question each one answers, LEI counts and each cohort's stated universe. Costs nothing and spends no credits. NOT FOR: individual applications or borrower data. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
run_cohort	Screen a ready-made public cohort by id (see list_cohorts). Behaves exactly like screen_counterparties but with a published LEI list, so nothing has to be uploaded. Without a license it returns the aggregate shape plus the first three rows; with a license key it returns the full cohort and an Evidence Brief, and spends one credit per covered LEI. NOT FOR: individual applications, lender recommendations or any misconduct conclusion. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.
get_conditional_door_map	Conditional surface for a lender: how its peer-adjusted denial gap changes across leverage bands inside identical published cells (state x loan amount x income x DTI x CLTV). Use when someone asks whether a lender is strict or lenient and the honest answer depends on the loan. RETURNS: the gap per CLTV band under two weightings, cell and decision counts, a within-lender permutation p-value, the flip_region where the sign changes, and the four robustness checks the pattern survived (bootstrap flip location, negative control on income bands, empirical-Bayes shrinkage, leave-one-state-out). NOT FOR: approval likelihood, policy cutoffs, deciding which lender will approve a file, or any individual application. The data locate a transition BETWEEN published bands and never at a point: point_threshold_available is always false. Report this as a band-level observational pattern, never as a lender policy. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.