Extracting data from rate locks
A lender fixes a mortgage rate for a set number of days and confirms it in writing, on whatever form the investor or housing agency uses. A correspondent lender locks every loan it sells on, and the confirmation is what the file is checked against. Large sellers handle thousands a month, arriving as portal downloads.
A secondary marketing desk needs the locked rate, the lock period, the expiry date and the program the loan was priced under. A loan origination system holds the loan to those terms, and a pipeline report counts what expires this week.
What makes rate locks hard
The terms fit one wide row
Program, amount, rate, price, lock date, period and expiry are set in eight narrow columns across a single line.
Inputs sit apart from the price
The credit score, the ratios and the occupancy that produced a rate are printed in a summary well below the rate itself.
Adjustments carry their reasons
Each price adjustment names what triggered it, so a final price is a base plus a list of separately labeled moves.
Parse, extract, verify
Parse
Parse turns the document into reading-order Markdown and a hierarchy of blocks (text, tables, figures), and grounds each block to its exact location on the page.
Extract
You define the fields in a JSON schema, and Extract returns their values in a consistent, structured format, each linked back to where it appears in the Markdown.
Verify
Each value links back to where it appears, down to the character range in the Markdown, and the page and bounding box in the source document, so every result can be verified.
One rate lock, end to end
About this document: WHEDA's own sample rate lock confirmation for its correspondent channel. One legal-size page: the lock itself in a single wide row, then pricing adjustments, the loan officer, and a two-column summary of borrower, property, loan, underwriting and mortgage insurance. Page 1 is featured. The borrower and property are placeholders.
Processing objective: The objective is to read the terms a desk holds the loan to, and the inputs that produced them. The rate sits in a row of eight narrow columns, and the credit score that priced it sits in the summary further down the page.
Demonstration fields: Reference number, Loan program, Locked rate, Lock expiration date, Middle credit score
The document
1 page
Page 1shown below
Where each value came from
reference_number
{
"value": "800203",
"page": 1,
"range": {
"end": 143,
"start": 123
},
"box": {
"xmin": 0.73165,
"ymin": 0.22396,
"xmax": 0.88156,
"ymax": 0.23279
}
}rate_lock.loan_program
{
"value": "HFA Preferred with MI Correspondent 30 year",
"page": 1,
"range": {
"end": 372,
"start": 329
},
"box": {
"xmin": 0.07216,
"ymin": 0.27766,
"xmax": 0.27638,
"ymax": 0.29996
}
}rate_lock.interest_rate_percent
{
"value": 4.625,
"page": 1,
"range": {
"end": 452,
"start": 409
},
"box": {
"xmin": 0.36008,
"ymin": 0.27766,
"xmax": 0.71952,
"ymax": 0.29995
}
}rate_lock.expiration_date
Printed on the page as 4/10/2014, normalized on extraction.
{
"value": "2014-04-10",
"page": 1,
"range": {
"end": 470,
"start": 461
},
"box": {
"xmin": 0.71941,
"ymin": 0.27766,
"xmax": 0.80505,
"ymax": 0.29994
}
}borrower.middle_credit_score
{
"value": 715,
"page": 1,
"range": {
"end": 1056,
"start": 1051
},
"box": {
"xmin": 0.34702,
"ymin": 0.47723,
"xmax": 0.50595,
"ymax": 0.48903
}
}How to try it yourself
Everything on this page is reproducible. The sample, its schema, and the Parse and Extract API outputs are published. You can repeat the same two calls from whichever surface you already build on.
- Claude Skilldescribe the job; the agent writes it
- CLIno code at all
- MCPexpose ADE docs as an agent tool
- REST APIany language
- Pythonlandingai-ade
- TypeScriptlandingai-ade
Documents into data
Accurate
The best accuracy on the difficult documents other tools get wrong, such as messy scans, complicated tables, checkboxes and handwriting.
Auditable
Every answer cites its sources. Values come back with an audit trail (page, bounding box, character range) so a human can review them.
Agentic
Applies multi-step reasoning, and plans before it acts. Handles documents it has never seen before and adapts to variations.
API-first
Composable APIs that drop into any tech stack and can be called from anywhere. Easy to build on and embed in your pipelines and products.
Affordable
Cost-competitive rates with complete price transparency. Spend controls before you run, and every job reports its exact cost.
Enterprise-grade
Built for production and scale from day one. Fully observable and monitored, running securely in our cloud or yours.
Built for high-volume, high-throughput production
- Zero data retention
- Priority rate limits
- Enhanced support
- VPC and on-prem deployment
Available on Team and Enterprise plans.
