MapAgent: A Hierarchical Agent for Geospatial Reasoning with Dynamic Map Tool Integration

TL;DR

MapAgent presents a hierarchical, multi-agent framework for geospatial reasoning that decouples planning from execution and introduces a Map-Tool Agent to orchestrate map APIs. By supporting parallel and sequential tool use through four specialized map tools, MapAgent overcomes tool inflation and capability limitations inherent in flat agent architectures. Empirical evaluations on MapEval-Textual, MapEval-API, MapEval-Visual, and MapQA show consistent gains over state-of-the-art baselines across diverse backbones, with notable improvements in multi-hop spatial reasoning and data-grounded answering. The approach offers robust, plug-and-play integration with real-world map services and demonstrates practical potential for automated, map-aware decision support.

Abstract

Agentic AI has significantly extended the capabilities of large language models (LLMs) by enabling complex reasoning and tool use. However, most existing frameworks are tailored to domains such as mathematics, coding, or web automation, and fall short on geospatial tasks that require spatial reasoning, multi-hop planning, and real-time map interaction. To address these challenges, we introduce MapAgent, a hierarchical multi-agent plug-and-play framework with customized toolsets and agentic scaffolds for map-integrated geospatial reasoning. Unlike existing flat agent-based approaches that treat tools uniformly-often overwhelming the LLM when handling similar but subtly different geospatial APIs-MapAgent decouples planning from execution. A high-level planner decomposes complex queries into subgoals, which are routed to specialized modules. For tool-heavy modules-such as map-based services-we then design a dedicated map-tool agent that efficiently orchestrates related APIs adaptively in parallel to effectively fetch geospatial data relevant for the query, while simpler modules (e.g., solution generation or answer extraction) operate without additional agent overhead. This hierarchical design reduces cognitive load, improves tool selection accuracy, and enables precise coordination across similar APIs. We evaluate MapAgent on four diverse geospatial benchmarks-MapEval-Textual, MapEval-API, MapEval-Visual, and MapQA-and demonstrate substantial gains over state-of-the-art tool-augmented and agentic baselines. We open-source our framwork at https://github.com/Hasebul/MapAgent.

Figures (9)

• Figure 1: Performance comparison across 4 geo-spatial benchmarks. MapAgent significantly outperforms others and achieves state-of-the-art (SOTA) performance.
• Figure 2: Overview of MapAgent. Given a user query (optionally with an image), the Planner Agent decomposes it into subtasks using the available Module Inventory and selects appropriate modules for each subtask. For tool-heavy modules, a dedicated agent (e.g., a Map-Tool Agent) manage interactions with the associated tools adaptively.
• Figure 3: Performance gain from API-based visuals.
• Figure 4: Performance gain from combining contextual text with API data over text alone.
• Figure 5: Cost analysis of MapAgent compared with Chameleon and Octotools