Skip to main content
When an AI assistant connected to Ploti finds the parcels that answer your question, it can show them on a Ploti map instead of in a long table. The map holds the records it found beside the layers the answer rests on, behind a link that opens in Ploti. It stays in your workspace (your team’s Ploti account), a teammate can open it, and it does not expire. The assistant does all of this for you. This page explains what it is doing, for when you want to steer it or build an agent that does the same. A connected chat assistant calls these steps as tools, named below. An assistant that uses only Ploti’s command-line tool, such as Muse, runs the same steps as ploti maps ... commands, shown in the CLI tab on the right. Every step needs a key or connection with the Maps permission. Reading data never needs a map.

​
Make a map

  • maps_create takes a title and, optionally, a place to center on. It returns the map’s id and a link to open it.
  • maps_list lists your maps, most recently used first, and maps_get returns one, so another conversation can keep working on the same map tomorrow.
  • maps_navigate moves the view you see to a place. It never limits what the assistant can read.
  • maps_datasets lists what the map offers at its current view.
A workspace may make 20 maps an hour.

​
Snapshots

A snapshot is a saved set of records from one layer, such as the 40 parcels that passed a screen. You can review it on the map or as a table. maps_snapshots_create picks its records in one of four ways: by their ids (up to 100,000), by conditions such as acreage over 5 in a place, from a table in the map’s worker, or as shapes no layer holds, such as a drawn study area. A snapshot can carry columns to show with each record, a legend and colors. maps_snapshots_delete removes one, unless a filter uses it.

​
Filters

A filter is a saved, live view of up to 12 entries. Each entry is a snapshot or a layer with conditions, with a title and a color. A layer entry is a rule, so it shows whatever matches each time it is opened. maps_filters_create saves one. maps_filters_update replaces its entries and saves a new version.

​
Show and share

  • maps_outputs_list lists what a map holds: snapshots, filters and files.
  • maps_outputs_show returns a short preview and a link to open a snapshot or filter in Ploti. Some assistants, such as Claude Desktop, also show it as an interactive map or table in the chat.
  • maps_outputs_share makes a public, view-only link that lasts 7 days, for someone outside your workspace.
  • maps_files_export saves a dataset’s rows as a CSV, GeoJSON or Parquet file on the map. It returns a download link that lasts 15 minutes and a lasting link to open the file in Ploti. It takes the same conditions as datasets_rows, plus the map’s map_id: exporting always needs a map.

​
Analysis in the map’s worker

Joins across datasets, statistics, distances to the nearest feature and scoring need the rows together in one place. For that, each map has a worker: a private database on Ploti’s servers, running DuckDB, a fast analytics database. Only assistants connected over MCP can use it.
  • maps_tables_load copies rows from a dataset into a worker table, with no row limit, and maps_tables_documents_load copies up to 20 documents into tables of their passages.
  • maps_tables_compute runs one read-only SQL SELECT over the worker’s tables and can save the result as a new table.
  • maps_tables_analyze runs a statistical or spatial method, such as a nearest-feature match or a spatial join; maps_tables_methods lists them.
  • maps_tables_list, maps_tables_drop and maps_tables_export (a file of up to 500 MB) manage the tables, and jobs_get waits for a long step.
A worker stops after 15 minutes idle and starts again with its tables when next used. After an hour unused, it is removed with its tables. Snapshots, filters and files stay. An assistant that uses only the CLI has no worker. It downloads the datasets and runs the same SQL in its own DuckDB.