MCP for Claude
42 task-oriented tools Claude can call directly. Most are 1:1 wrappers around REST endpoints — same shape, same data. 5 are composites that collapse multi-step workflows (sponsor-book analysis, drift verdicts, predicate screening, market movers).
Connect Claude
# claude.ai → Settings → Connectors → Add custom connector URL: https://api.fhalenderdata.com/mcp/ # OAuth is auto-discovered. claude.ai will pop up a sign-in # the first time you call a tool. Re-grant takes one click.
Add the connector once. Then ask diligence questions in plain English and Claude will pick the right tool from the catalog below — tool docstrings carry intent examples ("use when the user asks X") so selection is consistent across sessions.
Lender lookup & profile
Resolve a name or ID to the lender, then pull any view. The profile call accepts EITHER a HUD lender_id or a name fragment — name-search with multiple hits returns a disambiguation envelope so Claude can re-call with the chosen ID.
find_fha_lenderSearch lenders by name → ID + HQ + active branch count + CWT flag. Use when you only have a name and want a quick disambiguation list.
get_lender_profileAccepts ID or name. Identity, mortgagee type, LI/DE status, branch counts, principal/agent counts, AND folded-in DBAs. Returns multi-match envelope on ambiguous name.
get_lender_performanceHeadline CR + SDQ + origs + T/R/S split at any geo, with symmetric area-channel peer benchmark.
get_lender_historyUp to 13 quarter-ends of CR trend with full Total/Retail/Sponsored split per snapshot.
get_lender_branchesAll branches with address + phone + DBA + dates.
get_lender_branches_inBranches scoped to a geography (state / zip / county / msa).
find_lender_by_branch_idReverse lookup: branch ID → owning lender.
get_lender_relationshipsPrincipal + agent counterparty lists — who this lender sponsors and who sponsors them.
get_lender_hud_officesEvery HUD field office the lender operates in, with latest AND prior snapshot per office (T/R/S each). The right tool for 'where is this lender at CWT risk right now and is it getting worse'.
Lender — Direct Endorsement (T/R/S)
Total / Retail / Sponsored channel split. Where the sponsor-heavy vs retail-heavy story actually lives.
get_lender_de_breakdownT/R/S CR + loan counts + area-channel peer benchmark at one geo.
list_lender_de_geosEvery geo at one level where the lender has DE activity, sorted by volume, with T/R/S split. Use to see geographic distribution of the book.
compare_lenders_de_at_geoN lenders compared side-by-side with T/R/S at one geo.
Geographic discovery
Find lenders in a place. The dispatcher accepts every geo type and folds Metropolitan Divisions for the 11 OMB-split CBSAs (Atlanta/Chicago/NYC/etc). HUD field offices live in their own family because HUD enforcement is geo-scoped per office, 24 CFR 203.3(d).
find_lenders_in_geographyCanonical 'lenders in X' tool. Accepts geography_spec like 'state:CA', 'zip:90210', 'msa:12060', 'county:48201'.
find_lenders_in_hud_officeLender roster for one HUD office, sorted by volume. Accepts office NAME ('ATLANTA') or numeric CODE ('406').
Geographic aggregates
Single-geo summaries + time series. Works at state / county / msa / zip / tract / cd.
get_geo_summaryAggregate stats for one geo: originations, distinct lenders, SDQ rate.
get_geo_trendTime series of any metric at any geo.
get_county_sdq_summaryAll counties with originations + SDQ rate. Choropleth-shaped.
get_county_zip_breakdownZIPs in a county with per-ZIP SDQ rates.
get_hud_offices_sdq_summaryEvery HUD office with originations + SDQ. Choropleth-shaped.
Boundaries
Polygon + centroid data for in-tool map artifacts. State + county polygons aren't exposed — use the us-atlas CDN directly.
get_geo_boundarySingle GeoJSON Feature for one MSA or ZIP (~5–50KB).
get_geo_centroidsLightweight {code, name, lat, lng, bbox} per geo. For dot/scatter maps.
Geo name resolvers
Convert user-friendly names ('Atlanta', 'TX-23') to the codes other tools want.
find_msaName → CBSA code resolver. Matches HUD NW dropdown names.
find_congressional_district'TX-23' → CD code.
find_tract11-digit census tract code lookup.
Market analysis
National + cohort lens. Note the composite tools — these collapse multi-step workflows into single calls.
get_market_summaryLatest national snapshot with top-5 by volume + best/worst CR.
compare_lender_to_peersLender vs HUD volume-tier peer group (Mega/Large/Medium/Small/Micro).
compare_to_peer_setcompositeLender vs custom peer set (explicit list OR auto-pick by volume).
compare_periodsOne lender across two snapshot dates with deltas.
get_market_moverscompositeTop N lenders by QoQ CR swing — 'what changed last quarter'.
get_credit_watch_lendersLenders currently flagged on HUD Credit Watch (CWT).
get_recent_de_terminationsRecent Federal Register DE-termination notices with affected lenders.
screen_lenderscompositePredicate search ('lenders with CR 140-175, >5K origs, in TX or FL'). Replaces 'string together 5 tool calls' pattern.
flag_lender_driftcompositeStructured CR-drift verdict (ok / monitor / escalate). Avoids LLM trend-math errors on sparse series.
analyze_sponsor_bookcompositeEnd-to-end TPO/correspondent-risk diligence: sponsor's own DE + agent book health + structural verdict.
Servicers
Mirror of the lender-side surface for the servicing leg.
find_servicerSearch servicers by name + portfolio size + tier.
get_servicer_portfolioActive portfolio + 30/60/SDQ delinquency + foreclosure rate.
get_loss_mit_breakdownActive workout pipeline + lifetime claims by workout type + workout ratio.
get_top_servicersTop-N by portfolio size / sdq_rate / workout_ratio.
get_servicer_geo_concentrationState-level concentration of a servicer's portfolio + HHI.
get_servicer_tier_rankingHUD TRSII tier (1-4) + grade + score + Part-B reimbursement rate.
list_trsii_servicers_by_tierAll servicers at a TRSII tier (1-4) for the latest or named round.
Reference
Data freshness and snapshot metadata.
get_latest_snapshot_infoWhen the data was last updated + available historical quarters.
Ready to wire Claude in?
AI Edition includes MCP access with OAuth. Subscribe, then add the connector above.
See plans