MCP server exposing 4 tools for hmda.
This URL is a JSON-RPC 2.0 endpoint over HTTP. Issue POST requests with a JSON-RPC body. Browsers and search crawlers land here on GET.
POST https://gateway.pipeworx.io/hmda/mcp
Content-Type: application/json
{"jsonrpc":"2.0","id":1,"method":"tools/list"}hmda_lender_activity — What a specific mortgage lender did in a year — applications, originations, denials and dollar volume — from the federal Home Mortgage Disclosure Act register, optionally narrowed to a state or county. Identify the lender by its LEI (the 20-character Legal Entity Identifier HMDA files under); resolve_entity maps a company name to one. Covers every US mortgage application 2018-2025. Reports what lenders DISCLOSED, which is a regulatory filing rather than a market estimate.hmda_geography_activity — Mortgage lending in a place — how many applications were originated, denied or withdrawn in a county or state in a given year, with dollar volume, from the federal HMDA register. Break the answer down by applicant race, ethnicity, sex, loan purpose or loan type to see how outcomes differ across groups, which is the fair-lending question this data exists to make answerable. Covers 2018-2025.hmda_loan_records — Individual mortgage application records from the HMDA register — the loan-level rows behind the totals, with loan amount, action taken, rate spread, applicant income, lender LEI and census tract. Use when an aggregate is not enough and the individual applications matter. Filtered by geography and outcome; a large county returns tens of thousands of rows, so this caps what it returns and says when it truncated.hmda_lenders_in_county — Which lenders were most active in a county, ranked by loan count, from HMDA loan-level records for that county and year. Returns each lender LEI with its originations, denials and dollar volume, so "who lends here" and "who denies most here" are answerable. Bounded: it reads the county's records directly, so very large counties are sampled rather than fully ranked, and the response says so.Code samples (curl / TypeScript / one-click client install), schemas, and the live playground are on the pack page:
https://pipeworx.io/packs/hmda/
Pipeworx is an open MCP gateway connecting AI agents to live data. pipeworx.io