Graph Engine is a distributed, in-memory, large graph processing engine. It's a general-purpose computation engine that provides a unified declarative language for data modeling and message passing. It can be integrated with other system stacks via user-defined programming interfaces and RESTful interfaces.
"Trinity (Graph Engine) supports online query processing and offline analytics on large graphs," explained a Microsoft Research page about the project. "For online query processing, it leverages its fast graph exploration capability provided by the memory-based storage infrastructure. For offline graph analytics, it leverages the parallelism provided by the underlying scale-out distributed architecture."
Another Microsoft Research project, ProBase, was powered by Trinity. ProBase, according to the project page, has to do with natural-language/machine-language understanding and artificial intelligence. I wonder whether Graph Engine might also have anything to do with Microsoft's Cosmos massively-parallel computation and analysis service, which the company is working to commercialize.
Graph Engine includes a number of other features for distributed programming, including full Visual Studio/IntelliSense support; Language-Integrated Query and remote application deployment, control, monitoring and debugging. It also integrates with Azure, so that developers can create Graph Engine applications in VS and deploy to Azure.
Because it's seemingly still a Microsoft Research project, there are no guarantees on if and when Microsoft might take Graph Engine commercial. But there's a note on the preview-release page that hints this may be in the cards: "We will continue to provide routine maintenance and regular new releases, turning our research ideas into real-life functionality."
For those curious about the Trinity codename, I believe it is Matrix-related. One of the original graph database companies, Neo Technology, has a product called Neo4j. Neo's sidekick in the Matrix is named Trinity.