The ASINT API
Three surfaces: property valuation, vehicle valuation and the location index. One format, one authentication scheme, one data model underneath them.
What is available
The valuation model documentation is published and open — you can read it before we agree on anything.
Link Home
Valuation of apartments and houses. Inputs: district, mahalla, area, floor, room count, building type, finish condition.
Link Home documentation →Link Auto
Vehicle valuation. Inputs: make, model, year, mileage, body type, transmission, fuel, condition, region.
Link Auto documentation →GeoScore
The location index for a coordinate: a score for each of the nine subindexes on an absolute 0–100 scale. A batch mode exists for processing addresses in bulk.
GeoScore methodology →Request format
REST over JSON. The response always carries more than a number — an interval and a confidence level, because a valuation without them cannot be defended to a credit committee.
The full field list, accepted values and error codes are in the per-model documentation above.
POST /service/predict/home
Authorization: Bearer <token>
Content-Type: application/json
{
"district": "Yunusobod",
"area": 62,
"rooms": 3,
"floor": 4,
"total_floors": 9,
"qurilish_turi": "panel",
"renovation": "euro"
}Access and limits
Keys are issued against a use case: we look first at the volume and the segments you need.
Authentication
JWT. The key belongs to an organisation, not to a user.
Format
REST, JSON in and JSON out. A Swagger description is available on request.
Batch processing
Bulk address processing, asynchronous, with a callback on completion — for revaluing a whole collateral portfolio.
Limits
Depend on the plan. The rate limit is returned in the response headers.
An MCP server for AI agents
Alongside REST we expose an MCP server, so assistants and agents built on large language models can query valuation and location data directly, without a separate integration written for each tool.
The practical difference: an agent does not need someone to write a wrapper for it. It receives a description of the available tools and calls them itself. If you are building an internal assistant for a bank or an agency, this is the shortest path.
Request MCP accessHow it usually gets connected
The three scenarios people arrive with most often.
Pre-screening applications
Automated valuation at the front of the mortgage pipeline: some applications are filtered out or confirmed without sending an appraiser.
Collateral portfolio monitoring
Periodic revaluation of the whole portfolio in a batch request, instead of a manual quarterly re-appraisal.
Location risk in scoring
Risk subindexes as an input to the credit decision: a threshold on natural or technogenic hazard changes the insurance requirement or the LTV.
Need a key, or a pilot?
Describe the use case and the expected volume — we will propose a plan and give you test access