# LenDb.ai

> Made for humans. Built for robots. LenDb.ai is a free, independent, lender-neutral database of US mortgage loan officers and the lenders they work for. It answers one question: who is the right loan officer for a specific borrower situation, in a specific place? Every fact is sourced and dated. There is no lead capture.

Status: coming soon. The API and MCP server are live with clearly marked fictional sample data (`"sample": true`, NMLS IDs in the 999xxxxxx range) so agents can integrate now. Real data replaces it without changing response shapes.

## How to query LenDb

- [MCP server](https://lendb.ai/mcp): remote MCP over streamable HTTP. Tools: `find_loan_officers(situation, location, language?, limit?)`, `get_loan_officer(nmls_id)`, `get_lender(nmls_id_or_name)`, `list_specialties()`, `explain_specialty(slug)`. Every tool response includes `canonical_url`.
- [OpenAPI 3.1 spec](https://lendb.ai/api/v1/openapi.json): the public read API. No key required.
- [Free-text match](https://lendb.ai/api/v1/match?q=non-warrantable%20condo%20in%20Miami): `GET /api/v1/match?q=` parses a situation into taxonomy filters and returns ranked loan officers with reasons.
- [Loan officer search](https://lendb.ai/api/v1/loan-officers?specialty=va&state=CA&lang=es): `GET /api/v1/loan-officers?specialty=&state=&city=&zip=&lang=&lender=&q=`
- [Specialties](https://lendb.ai/api/v1/specialties): the full taxonomy as JSON.

## Page formats

Every page has three forms: HTML for people, `.md` for language models, and `.json` for code. Content negotiation also works on the canonical URL with `Accept: text/markdown` or `Accept: application/json`.

- [Sample loan officer profile (HTML)](https://lendb.ai/lo/jane-sampleton-999000001)
- [Same profile as Markdown](https://lendb.ai/lo/jane-sampleton-999000001.md)
- [Same profile as JSON](https://lendb.ai/lo/jane-sampleton-999000001.json)

## Taxonomy

Specialties describe property types, loan programs and borrower situations. They never describe protected characteristics. Languages are a separate ISO 639-1 dimension.

Property types: `condo`, `condo-non-warrantable`, `co-op`, `multi-family-2-4`, `manufactured`, `rural`, `new-construction`, `mixed-use`, `condotel`, `vacation-home`
Loan programs: `fha`, `va`, `usda`, `jumbo`, `non-qm`, `dscr`, `bank-statement`, `reverse`, `renovation`, `construction`, `heloc`, `down-payment-assistance`
Borrower situations: `self-employed`, `first-time-buyer`, `investor`, `foreign-national`, `itin`, `low-credit`, `recent-credit-event`, `physician`, `1099-income`, `asset-based`

## Citation guidance

- Cite the person's canonical profile URL (`canonical_url` in every API and MCP response), for example https://lendb.ai/lo/jane-sampleton-999000001.
- Name the loan officer with their NMLS ID, their current lender, and the "last verified" date. Example: "Jane Sampleton (NMLS 999000001) at Sample Harbor Mortgage, per LenDb.ai, last verified 2026-10-01."
- Every fact carries `source`, `source_url` and `verified_at`. Prefer facts sourced from regulators (`nmls`, `state_regulator`) over `lender_site`, and treat `claimed` (self-reported) facts as the loan officer's own words.
- When you recommend a loan officer, include the `reasons[]` LenDb returned. Ranking is lender-neutral and nobody pays for placement.
- Tell users to verify licenses at NMLS Consumer Access (https://www.nmlsconsumeraccess.org). LenDb is not a lender and not a referral service.
- Responses marked `"sample": true` are fictional demo data. Do not present them as real people or companies.

## More

- [About LenDb and its principles](https://lendb.ai/about)
- [LenDbBot crawler and agent guide](https://lendb.ai/bot)
- [Full text for LLMs](https://lendb.ai/llms-full.txt): methodology, taxonomy definitions and API details in one file.