count_by_city
ChatGPTCount parcels for ANY city-scoped question using an exact summary view. Always use this tool instead of count_properties when a count is scoped to a city. It returns an exact precomputed total plus the authoritative ZIP codes represented by that city in the parcel data. Never infer or assume a city's ZIPs from general knowledge. Matching uses the normalized uppercase city_key, so inconsistent source casing does not affect the result. If the city is absent, the response is empty and explicitly tells the caller to check spelling; never substitute guessed ZIP codes or a similarly named city.
count_properties
ChatGPTCount parcels matching supported filters — e.g. county, ZIP, or subdivision. The city filter is unsupported for counting because city counts scan 10.8M rows and either time out or produce unreliable estimates (observed 31x low). Use count_by_city for any city-scoped count; it returns an exact precomputed total and the authoritative ZIP codes represented by that city in parcel data. For supported filters, attempts an exact count first. If the database statement times out, retries with a planner estimate and marks the response as approximate. At least one filter is required. Combine filters to narrow (e.g. county + subdivision, or zipcode + use_code). Note: counting by subdivision (matched against the legal description) is only supported together with a county — a statewide subdivision-name scan is too slow. County, ZIP, owner, and value/year counts work on their own.
get_hoa_detail
ChatGPTGet the full record for one HOA / community association by its doc number. Returns the association, its management company, the community association manager, and how many parcels are linked to it.
get_hoa_for_parcel
ChatGPTGet the HOA / community association (and its manager) governing a parcel. Direct links exist only for HOA-owned parcels (common areas), because linkage is by owner name. For an individual home with no direct link, this tool (with infer=True, default) infers the HOA from association-owned parcels in the same appraiser neighborhood — clearly tagged as inferred. A parcel may return more than one association (e.g. a master + sub-association).
get_property
ChatGPTGet the FULL record for a single parcel by object_id. Returns every property field (identification, owner, fiduciary, all just/assessed/taxable values, building and land characteristics, legal description, roof estimates, geo centroid) plus a clean sale_history array and human-readable county_name / dor_uc_desc.
get_property_by_parcel_id
ChatGPTGet the FULL record for a parcel by its parcel_id (exact match). Same complete output as get_property. If a parcel_id is not unique statewide, pass county to disambiguate; if multiple still match, the matches are returned for you to pick an object_id.
get_sale_history
ChatGPTGet the sale history for a parcel (most recent sale first). Returns the property's recorded sales with price, year/month, OR book/page, clerk number, and qualification/vacant-improved codes.
list_columns
ChatGPTList every queryable column of the parcels table with its type. Use this to discover valid column names (and types) for query_parcels and for the use_code/filter arguments elsewhere.
query_parcels
ChatGPTQuery parcels by ANY column(s) — a flexible filter over the whole table. Use this when the fixed search tools don't cover the field you need (e.g. township/range/section, neighborhood code, homestead values, roof area, market area, fiduciary, etc.). Call list_columns to see queryable columns. Each filter is an object: {"column": <name>, "op": <operator>, "value": <v>}. Operators: eq, neq, gt, gte, lt, lte (comparisons); like, ilike (text — the value is treated as a 'contains' match unless you include * wildcards); in (value = list); is (value = null/true/false). All filters are AND-ed. Tip: pair a broad text filter with an indexed one (county/city) for speed — scans over unindexed columns across 10.8M rows can be slow.
search_by_address
ChatGPTSearch parcels by physical street address (case-insensitive partial match). Returns a summary list. Use the object_id of a result with get_property to retrieve the full record.
search_by_filters
ChatGPTSearch parcels by multiple criteria (at least one filter required). Results are ordered by just value (highest first). Use get_property for the full record of any result.
search_by_owner
ChatGPTSearch the public Florida assessment roll by owner of record (partial match). Intended for entity and portfolio research: which parcels a company, trust, or estate holds, and how that holding spreads across counties. Owner names are as printed on the Florida Department of Revenue tax roll, a public record that carries no contact information. Returns a summary list; use get_property for the full record of any result.
search_by_parcel_id
ChatGPTLook up parcels by parcel ID (exact match or prefix search). For a single known parcel ID, prefer get_property_by_parcel_id which returns the full record directly.
search_hoa
ChatGPTSearch HOA / community associations by name, city, ZIP, or type. At least one filter is required. Returns association cards including the management company and manager where known. Use get_hoa_detail for a single HOA's full record (managers + how many parcels it covers).
search_nearby
ChatGPTFind parcels near a geographic point (by parcel centroid), sorted by distance.
count_by_city
Claudecount_properties
Claudeget_hoa_detail
Claudeget_hoa_for_parcel
Claudeget_property
Claudeget_property_by_parcel_id
Claudeget_sale_history
Claudelist_columns
Claudequery_parcels
Claudesearch_by_address
Claudesearch_by_filters
Claudesearch_by_owner
Claudesearch_by_parcel_id
Claudesearch_hoa
Claudesearch_nearby
Claude