Mapbox's agentic mapping engine and new services give AI applications, agents, and developer tools the location-grounded reasoning to deliver more accurate and actionable experiences.
SAN FRANCISCO, Sept. 17, 2026 /PRNewswire/ -- Mapbox, which provides location infrastructure for AI, today introduced a new set of services and capabilities designed to give AI applications, agents and developer tools a more accurate understanding of the physical world. Mapbox CEO Peter Sirota announced the new capabilities during the opening keynote at Mapbox BUILD.
AI struggles to understand location and often fails to provide useful location information, producing generic recommendations that ignore real surroundings and intent.
Mapbox closes the location gap by giving applications, agents, and developers access to current, precise location infrastructure built on an agentic mapping engine. The location infrastructure enables AI applications and agents to understand locations, proximity, routes, and changing real-world conditions so they can provide more relevant answers and take more useful actions.
"AI can explain almost anything, but it doesn't deeply understand physical location," said Peter Sirota, Mapbox CEO. "Mapbox delivers location infrastructure for AI that enables AI models, apps, and agents to move beyond generic responses and deliver recommendations and actions that reflect a user's surroundings, circumstances, and intent."
Location Infrastructure Built for AI
At the center of the announcement is the Mapbox agentic mapping engine, which uses autonomous agents to continuously process live inputs and anonymized movement data from over 45,000 applications to help Mapbox keep roads, traffic conditions, and map data current in near real time.
Mapbox released the public preview of the new Mapbox Places API, which gives developers access to rich, structured information about points of interest. Powered by the Mapbox agentic mapping engine, the API combines core business information with geographic details such as building footprints, operating hours, entrances, and visitation patterns, giving developers a comprehensive spatial dataset spanning over 250 million points of interest globally.
Mapbox also announced Mapbox Traffic 2.0, the next-generation traffic engine that delivers accurate estimated time of arrival (ETA) for 98 percent of trips, incorporates congestion at major maneuvers, and forecasts traffic up to 2.5 hours ahead. Powered by AI models that learn by comparing generated estimates with actual arrival times, Traffic 2.0 leverages the closed learning loop of the Mapbox agentic mapping engine that continuously incorporates live movement and historical patterns. The result is more precise and reliable traffic information for drivers and logistics systems.
"Location isn't just another data point for an AI model. It's the bridge between digital intelligence and physical action," Sirota added. "As AI transitions from generating answers to executing complex real-world workflows, Mapbox provides the foundational spatial engine that ensures every model response and automated action is grounded in the physical world."
Location infrastructure for applications
- Mapbox Natural Language Queries Search: An enhancement to the Mapbox Search Box API that interprets detailed conversational queries, such as "coffee shops near me with fast Wi-Fi and a quiet workspace," and translates them into relevant location results.
- Mapbox AI Mode for Static Images API: Building on the existing Mapbox Static Images API, AI Mode, restyles that image at request time with a prompt parameter - such as turning the same map into a summer afternoon full of tourists, or blanketing Paris in snow it rarely sees — all grounded in the real roads and buildings of the basemap.
Location infrastructure for agents
- Agent Toolkit for Maps and Navigation SDK: Lets agents read and modify context across more than 35 mapping and navigation controls: a driver can use voice commands to add a stop to a route or check the traffic impact on an upcoming meeting, and a chat agent can add points, control the map view, or get route summaries.
- Agent Integration with Notion Integration: Through a new Mapbox integration, Notion Agents can now use Mapbox tools to automate spatial workflows inside workspace documents — cleaning and standardizing customer addresses, tracking deliveries, coordinating field operations.
Location infrastructure for Builders
- Mapbox CLI: Lets developers and AI coding agents manage and deploy Mapbox services from the terminal, replacing manual work in the web console with command-line workflows.
- Demo Token: Enables developers and coding agents to start building immediately with just a verified email address. This allows teams to safely evaluate the platform and build location experiences without the risk of unexpected charges.
- Figma MCP connector: The new Mapbox connector for Figma places real maps in Figma designs, without screenshots or developer handoff. The connector allows a designer to describe a map in Figma's chat, such as "a map of central Copenhagen, dark theme", and the connector renders it at the requested location, zoom, style, and time of day, with routes, pins, and areas layered on top.
Developers can learn more by visiting: www.mapbox.com
About MapboxAbout Mapbox
Mapbox is the location platform for innovators. Headquartered in San Francisco, Mapbox provides developers and businesses with cloud-based APIs, SDKs, and data services for maps, navigation, and place and address search.
Founded in 2010, Mapbox leads development of the modern geospatial stack. Trusted by automakers, logistics providers, retailers, and app developers, Mapbox delivers scalable, highly customizable geospatial solutions that power billions of sophisticated location experiences worldwide. The platform is known for its built-in flexibility, elegant design, enterprise-grade security, strict data privacy, and developer-first documentation.
Learn more at www.mapbox.com.
SOURCE Mapbox
Share this article