Skip to content
supercarto
Esc
↑↓navigate↵open⌘Jpreview
On this page

Introduction

supercarto gives an AI agent a map of any location on Earth as a token-budgeted topological graph — one call, N tokens, measured on output.

supercarto gives an AI agent a map of any location on Earth as a token-budgeted topological graph: streets as edges, crossings as junctions, indoor floors and levels as separate layers, with weather, terrain, daylight and transit riding along. Everything the agent reads is measured in tokens, so one call fits a budget — and the budget is re-measured on the emitted output, not estimated on the input.

A give and a take

One file in, both numbers out: examples/synthetic-dense.geojson is 272,993 bytes (~68,000 tokens) raw and compiles at a 1024-token budget to about 890 tokens — 26 nodes (10 named) on 34 edges, ~99% smaller. What the agent reads:

map:
  center: "37.7749, -122.4194"
  source: geojson
  note: Simplified for reasoning; call expand_feature for exact geometry.
nodes:
  n530: {type: poi, name: Bank 0, tags: [bank]}
  n531: {type: poi, name: Cafe 1, tags: [cafe]}
  n673: {type: transit, name: Station 0, tags: [subway_entrance]}
edges:
  n515->n516: {dist: 80m, dir: east, path: road}
omitted:
  - {layer: building, count: 259}
  - {layer: poi, count: 141}
  - {layer: road, count: 441}
tools:
  - expand_feature(id)

It is standard YAML in flow style, using only constructs a model has read billions of times. The wire format is standard YAML, optimized for machines; the compiler, not the model, does the translation from raw geodata.

Install

npm install supercarto
pnpm add supercarto
yarn add supercarto
bun add supercarto
nub add supercarto
aube add supercarto

The quick start has the two-minute version for Node, Python, Claude Desktop, and the hosted endpoint.

  • The budget is a real budget — how output is ranked, cut, kept connected, and re-measured.
  • Honesty rules — the refusal and degradation semantics that run through every tool.
  • Accuracy — verified distances across latitudes.
  • Transit schedules — stops, lines, departure boards, live delays, and a certified ride plan.
  • FAQ — the questions every evaluation starts with.
  • Interfaces — MCP, CLI, HTTP, Python client.

Last updated on

Was this page helpful?