POST/ask

Ask About Lender Criteria

Query your processed cases and lender criteria in plain language.

Who is this for

Mortgage Broker / Adviser. You need to quickly check whether a case meets a specific lender's criteria without reading 50-page policy documents. Instead of searching PDFs manually, you ask a question and get a sourced answer in seconds.

What you'll do

1
Use the /ask verb to submit a natural-language question about lender criteria
2
Receive a structured answer with source citations from lender policy documents
3
Verify the answer against the cited sources
4
Optionally ask follow-up questions in the same conversation thread

This uses /ask, not /check

The /ask verb is a conversational query interface. It does not run compliance checks or score documents. Use it to look up criteria, ask questions about policy, and research lender requirements before preparing a case.

The API call

Ask a plain-language question with a cluster context to scope the knowledge base.

lender_query.py

python

from curvestone import Agent

agent = Agent()  # reads CURVESTONE_API_KEY from env

result = agent.ask(
    question="Does NatWest accept debt consolidation on remortgages over 85% LTV?",
    context={"cluster": "mortgage"},
)

print(f"Answer:     {result.answer}")
print(f"Confidence: {result.confidence}")
print(f"Sources:    {len(result.sources)}")
print(f"Cost:       {result.cost}")

The response

A structured answer with source citations and a confidence score.

response.json

json

1{
2  "id": "ask_9xKm2pRtV4",
3  "type": "ask",
4  "answer": "No. NatWest does not accept debt consolidation on remortgage applications where the LTV exceeds 85%. Their current policy (updated January 2026) caps debt consolidation remortgages at 80% LTV. For applications between 80-85% LTV, debt consolidation is permitted only with additional underwriter approval.",
5  "sources": [
6    {
7      "name": "NatWest Mortgage Policy v2026.1",
8      "section": "3.4.2 Debt Consolidation"
9    },
10    {
11      "name": "NatWest Broker Criteria Guide",
12      "section": "Remortgage — Special Conditions"
13    }
14  ],
15  "confidence": 0.94,
16  "cost": "£0.05"
17}

What happens

Conversational queries against the knowledge base

The /ask verb lets you query the platform in natural language. The agent searches across ingested lender policy documents, criteria guides, and regulatory references to produce a direct answer. You do not need to know which document contains the information — the agent finds it for you.

Answers include source citations

Every answer includes a sources array listing the document name and section that the answer was derived from. This lets you verify the answer against the original policy document and gives you an audit trail if a client or compliance officer asks where the information came from.

Low cost, high throughput

Each /ask turn costs £0.05 — much cheaper than running a full compliance check. This makes it practical to ask multiple questions while preparing a case, checking eligibility across different lenders, or verifying edge cases in lender policy.

Follow-up questions in the same thread

You can continue the conversation by passing the thread_id from a previous response. The agent retains context from earlier turns, so you can refine your query or ask related follow-up questions without repeating yourself.

0.90+High confidence. Well-sourced answer.

0.70 - 0.89Moderate. Verify with source documents.

Below 0.70Low confidence. Manual verification required.