Tool reference
The tools are grouped by topic. Query Ploti directly explains the reads in full. Data Feeds, the subscription that keeps a copy of the data in your own system, has no tools here.Find datasets
See Find datasets.Read rows
See Read rows.Documents and collections
See Documents and collections.Municipal codes
See Municipal codes.Maps
See Show results on a map.Analysis on a map
See Show results on a map.Knowledge graph
Ploti’s knowledge graph links the parcels, people, companies, addresses and recorded events found across the datasets in your coverage areas. Each answer names the source row behind it. Each entity has a public id that the other graph tools take.Research requests
See Research requests.Watch for changes
See Watch datasets.Maps and boundaries
Reading needs no map. The read tools name a place and cover everything in your coverage areas, the counties your plan covers, wherever any map is looking. A connection reads for one Ploti workspace only. A map is what you open in Ploti.maps_create makes one and returns its id, the same id the Ploti app opens. The tools that save outputs, move the view or run analysis take that id. A map does not expire: ask for it tomorrow from another conversation, and the assistant finds it with maps_list and keeps working on it.
Each map has a private worker for heavy analysis. A worker stops after 15 minutes idle and restarts with its tables when next used. After an hour unused, it is removed with its tables. The next call on the map starts a fresh worker, and maps_tables_load loads the tables again. Snapshots, filters and files saved on the map stay in the workspace either way.
Analysis and exports
Most questions need onlydatasets_rows and datasets_aggregate, which answer with a page of rows or a total at a time. Joins across layers, statistics and scoring move to a map’s worker. maps_tables_load copies in what is needed, and maps_tables_compute and maps_tables_analyze work on it there. Every worker reply says how much memory and disk are in use, so the assistant can work in smaller steps as it nears a limit.
maps_files_export and maps_tables_export save rows in the workspace as a CSV, GeoJSON or Parquet file of up to 500 MB. Larger exports are split into several files. Each export returns two links: a download link that needs no sign-in and lasts 15 minutes, which an assistant that runs code can fetch, and a Ploti link for you. Coding agents can also download a dataset’s rows as GeoParquet files with the CLI or the Python library.
Interactive outputs
maps_outputs_show always returns a short text preview and a Ploti link. Some assistants, such as Claude Desktop, also show an interactive map or table in the chat. A map and its comparison table can appear as separate outputs. The assistant can page through the preview or narrow it instead of putting a whole dataset into the conversation.
Use Filter rows to find entries in a table, Export CSV to download the result, or Open in Ploti to continue reviewing it in your workspace.

Review parcel attributes in Claude Desktop using the table's filter, Export CSV, and Open in Ploti controls.
Requesting data
When Ploti does not have the data a question needs, the assistant files a research request with your question. Ploti checks which data it needs, finds the official sources, and collects what is missing for the places in your coverage areas. That usually takes hours to days. Ploti then runs the research and emails you the answer. The assistant can check on it withrequests_get and requests_data_list. A delivered dataset can be searched and read like any other.
What the assistant can save
- Snapshots: fixed sets of places that can be reviewed on a map or in a table
- Filters: live layers on the Ploti map
- Files: CSV, GeoJSON and Parquet exports with authenticated download links