project · 2025-2026

Map SDK agent toolkit

Contributed to an agent toolkit that lets LLM applications interact with mapping capabilities through structured, typed tools.

#agentic #vercel-ai-sdk #typescript #multi-provider-llm #contributor

The problem

Conversational mapping experiences need more than a chat box: the application must translate intent into safe, structured map actions such as search, routing, and viewport updates.

Approach

Contributed to a headless toolkit that exposes mapping actions as typed agent tools.
Focused on predictable tool contracts and clear separation between language interpretation and deterministic map behaviour.
Made the interface compatible with a range of LLM providers and application shells.

How it works

The scripted demo shows the interaction model: a user expresses an intent, structured tools return mapping results, and the interface updates deterministically. It uses public example locations and no live services.

chat1 / 6

userShow me well-rated cafes in Berlin Mitte.

Demo basemap: open-source CartoDB / OpenStreetMap tiles, no API key required. The demo uses public tiles so the portfolio remains self-contained and key-free.

Scripted, not live. Click next → to step through the dialogue. Each agent message that has location output also re-frames the map and drops the relevant markers / route.

Outcome

The toolkit offers a reusable building block for conversational map experiences without coupling product teams to a particular UI.

← all projects