Add an MCP tool that returns a rendered resource map image

[ewlarson]

Summary

Expose the existing static-map capability through the MCP server so an AI client can request a map for a GeoBTAA resource and receive an actual image result rather than only a URL or textual geometry description.

The first implementation should reuse the current static-map service and cache behind GET /api/v1/static-maps/{resource_id}. It should not introduce a parallel rendering pipeline.

Proposed tool

render_resource_map(resource_id, width?, height?, format?) -> image

The final tool name and schema may follow the existing MCP naming conventions.

Behavior

• Resolve the requested resource using its canonical GeoBTAA identifier.
• Render or retrieve the cached static map for the resource geometry.
• Return the image through MCP as image content with the correct MIME type.
• Include compact textual context alongside the image: resource ID, title, geographic extent, canonical resource URL, and map endpoint URL.
• Return a useful structured error when the resource does not exist, has no usable geometry, cannot be rendered, or exceeds configured limits.
• Preserve the existing static-map fallback/placeholder behavior only when it communicates a real no-geometry state; do not present a placeholder as a successfully rendered geographic map.

Acceptance criteria

• A new MCP tool is discoverable through tools/list with a clear description and JSON schema.
• Calling the tool for a resource with valid geometry returns valid image content and the correct MIME type.
• The implementation delegates rendering and caching to the existing static-map service.
• The response includes enough text metadata for clients that do not display MCP image content.
• Unknown IDs, missing geometry, renderer failures, timeouts, and unsupported options produce stable, documented errors.
• Width, height, and output format are constrained to safe supported values with sensible defaults.
• Authorization and visibility rules are applied before returning metadata or imagery.
• Unit and MCP integration tests cover successful, cached, missing-geometry, unauthorized, invalid-input, and renderer-failure paths.
• The MCP documentation includes example requests and describes the returned content blocks.

Implementation context

Relevant existing components include:

• backend/app/services/mcp_service.py
• backend/app/tasks/static_maps.py
• GET /api/v1/static-maps/{resource_id}
• GET /api/v1/static-maps/{resource_id}/geometry
• backend/tests/services/test_mcp_service.py
• backend/tests/api/v1/test_static_map_endpoints.py