Vexcel is breaking new ground in geospatial AI access. On June 30, 2026, the aerial imagery and geodata giant announced the Vexcel Model Context Protocol (MCP), a connector that pipes its vast library of high-resolution aerial photos, 3D models, and geographic information directly into AI assistants like ChatGPT, Claude, and Microsoft Copilot. The move targets Windows-based IT teams, city planners, and enterprise users who increasingly lean on AI for spatial analysis and decision-making.

Instead of forcing users to juggle separate GIS platforms and AI chats, Vexcel MCP lets an IT analyst type a natural-language question inside Microsoft Teams or a desktop Copilot window and receive up-to-date, real-world geographic answers. “Where are the least obstructed rooftops for solar panels in this zip code?” or “Show me flood-risk parcels within 500 meters of recent storm damage” suddenly become simple queries that return precise, source-attributed results.

The Data Behind the Protocol

Vexcel operates one of the world’s largest collections of vertical and oblique aerial imagery. Its flights cover millions of square kilometers annually, producing orthorectified basemaps, digital surface models, and true-ortho mosaics with resolutions down to 3 cm per pixel. The company already serves global insurance carriers, telecoms, and government agencies through its Viewer and API products.

What the MCP connector adds is a native language interface. By implementing the Model Context Protocol—an open standard originally incubated by Anthropic and later adopted across the AI industry—Vexcel exposes its entire catalog through a structured, AI-ready endpoint. That lets any MCP-compatible assistant fetch, filter, and interpret live geospatial data without brittle custom integrations.

Supported Assistants and Windows Integration

Vexcel explicitly names ChatGPT, Anthropic’s Claude, and Microsoft Copilot as launch partners. For Windows-centric organizations, the Copilot tie-in is the headline grabber. Copilot already lives inside Edge, the Windows taskbar, Microsoft 365 apps, and Teams; with MCP enabled, an IT manager can stay in the flow of work while pulling parcel boundaries, elevation profiles, or change-detection overlays into a conversation.

Behind the scenes, the connector registers itself as an MCP server. When a user asks a geography-related question, the AI reaches out to Vexcel’s APIs, retrieves the relevant geotiff, vector tile, or metadata, and then weaves the answer into its response—complete with attributes like capture date, sensor type, and confidence score. The exchange happens over HTTPS, secured by OAuth 2.0, and complies with SOC 2 and ISO 27001 standards, according to Vexcel’s preliminary documentation.

Why Windows IT Teams Should Care

Microsoft has been steadily embedding Copilot into enterprise workflows. GitHub Copilot for code, M365 Copilot for documents and email, and Teams Copilot for meetings are already established. Adding geospatial context gives IT and operations teams a Swiss Army knife for physical-world problems.

Consider a municipality running its entire back office on Windows 11 and Azure. Their IT staff can now use Copilot to:
- Pre-screen thousands of properties for zoning compliance during a building-permit surge.
- Correlate road-surface degradation with recent aerial change-detection layers to prioritize maintenance.
- Instantly verify insurance claim damage by pulling pre-event and post-event imagery without leaving the claims-adjuster app.

Because the MCP server is hosted inside Vexcel’s cloud or, for sensitive workloads, can be deployed in a customer’s own Azure tenant, data never has to leave a controlled environment. Bandwidth costs plummet, too, because the AI only fetches the necessary tiles, not entire state-wide mosaics.

How It Works Under the Hood

The Model Context Protocol defines a schema for tools, resources, and prompts. Vexcel’s server registers several tool endpoints:
- find_imagery: searches by bounding box, date range, and resolution.
- get_3d_model: retrieves a photogrammetric mesh or point cloud.
- change_detection: compares two dates and returns a difference raster.
- attribute_query: retrieves vector attributes like ownership, land use, or assessed value (where licensed).

When an AI model detects intent, it formulates a tool call with JSON arguments. The MCP server executes the call against its image catalog and returns a structured response. The AI then synthesizes that into human-readable text, often including a rendered snippet of the orthoimage or a link to a full-resolution viewer. For Microsoft Copilot, Vexcel is also leveraging the Copilot extensibility framework, so the connector appears as a plugin in the Copilot store, one-click installable by IT admins.

Security and Compliance

Geospatial data is inherently sensitive—it can reveal troop movements, critical infrastructure, and private property details. Vexcel’s connector mirrors its existing security posture. Administrators define geofences (e.g., “only allow queries within this county boundary”) and can mask licensed third-party parcels. Every query is logged for audit, and content is encrypted in transit and at rest. Vexcel says it has completed a third-party penetration test and holds an active FedRAMP Tailored authorization, broadening its appeal to U.S. federal agencies that already run Windows-based desktops under DISA guidelines.

Real-World Pilot Projects

Three early adopters shared limited details under embargo. A major European insurer is piloting the connector with Claude inside a Windows virtual desktop environment to automate underwriting for hail claims. By asking Claude for “all roofs with visible impact craters in the last 48 hours,” adjusters triage claims 60% faster than using a traditional GIS portal.

On the public-sector side, a Midwestern U.S. county is feeding Copilot with Vexcel’s 2026 leaf-off orthoimagery to locate potential water-main leaks indicated by abnormal soil moisture patterns—an analysis that previously required two GIS specialists and a week of manual digitizing. The IT director, who requested anonymity until the county’s official release, said, “This puts geospatial intelligence in the hands of every frontline worker without teaching them QGIS.”

The Competitive Battlefield

Vexcel isn’t the first to bridge geodata and large language models. Esri has been integrating its ArcGIS software with AI through its own Assistant and Copilot connectors. Google Earth Engine offers Python-based LLM tool calls. Hexagon and Maxar also have experimental plugins. But Vexcel’s advantage is its native MCP implementation, which sidesteps vendor lock-in. Because MCP is open, any future AI assistant that supports the protocol—Gemini, Cohere Command, or open-source models—can instantly onboard to Vexcel’s server with no recoding.

For Windows IT shops, the open nature means they aren’t forced to adopt a single AI stack. A company could use Claude internally while partners use ChatGPT, both pulling from the same Vexcel MCP endpoint with identical data fidelity.

Licensing and Pricing

Vexcel says it will offer the MCP connector as an add-on to existing Image Viewer subscriptions starting at $2,500 per ten-power-user seat per year. Enterprise agreements that cover unlimited queries across an organization will start at $80,000 annually, with volume discounts for multi-year terms. A free trial with a 1 km² sample area will be available later in Q3 2026. The pricing, while enterprise-focused, reflects the high cost of maintaining daily aerial capture and compute-intensive photogrammetry pipelines.

Potential Pitfalls

Hallucination remains a risk. Vexcel’s connector outputs ground-truth data, but the AI model that interprets it can still misattribute facts or misrepresent confidence levels. In early testing, one model reported that a building was “constructed in 1942,” a date inferred from a nearby historical plaque that appeared in the imagery metadata; the actual build date was 1998. Vexcel mitigates this by encouraging prompt engineering that enforces a strict data-only response format and by displaying source images inline so users can visually verify.

Data freshness is another concern. Vexcel captures most urban areas once or twice per year, but disasters and construction move faster. Users will need to understand the temporal gap. The connector reports image capture date prominently, but it’s still on the human to decide if two-month-old imagery is sufficient.

The Road Ahead

Vexcel’s product roadmap teases deeper Copilot integration later in 2026: a natural-language 3D model generator that can create a textured mesh of a city block on command and embed it in a PowerPoint slide. The team is also working with Microsoft on a Graph-connected AI agent that can pull floor plans from OneDrive alongside Vexcel’s exterior imagery, giving facilities managers a complete digital twin of a property.

For the broader Windows ecosystem, this is another sign that AI is graduating from text generators to multimodal problem-solvers that understand space and time. IT leaders who start experimenting with geospatial AI now will be the ones who build competitive advantages in logistics, insurance, utilities, and government services. Vexcel’s MCP connector might just be the on-ramp they need.