AI and the SDK
The SDK brings AI in at two levels: for the developers building with it, and for the users of the applications they build.
| What it does | Who benefits | |
|---|---|---|
| Coding skill | Loads SDK knowledge into your coding agent | Developers building with the SDK |
| Agent Toolkit plugin | Lets users drive the map in natural language | Users of your application |
For developers: the coding skill
The tomtom-maps-sdk-js skill teaches AI coding agents such as Claude Code, Cursor, GitHub Copilot and Windsurf the SDK's module patterns, [longitude, latitude] coordinates, service APIs and pitfalls. Install it once in your project:
npx skills add tomtom-international/maps-sdk-js --skill tomtom-maps-sdk-js tomtom-maps-sdk-js-feedback
The optional tomtom-maps-sdk-js-feedback skill lets the agent report SDK problems it hits to TomTom, with your consent. Installing the skill also covers the skill of an older SDK version. AI coding with the SDK lists the topics the skill covers, how to invoke them and what a report holds.
For end users: conversational map control
The @tomtom-org/maps-sdk-plugin-agent-toolkit plugin turns a TomTom map into something a user can talk to: "Route from Berlin to Munich avoiding tolls", "Show coffee shops near me", "What's the traffic like on this route?". The agent does the geocoding, routing and map display.
- Headless: you bring the chat UI and the LLM, from any Vercel AI SDK provider.
- Tools built in for location, routing, places, traffic, reachable ranges and direct MapLibre access, with an intent classifier that keeps each prompt small.
import { createMapAgent } from '@tomtom-org/maps-sdk-plugin-agent-toolkit';
import { openai } from '@ai-sdk/openai';
const agent = createMapAgent(map, { model: openai('gpt-4o') });
The Agent Toolkit overview wires it to a chat UI.