Microsoft and Google implement new strategies for AI agent context management

Major technology providers Microsoft and Google are implementing different strategies to manage context in AI agent development to address rising operational costs and data management challenges.

Microsoft Foundry is utilizing context engineering to reduce AI agent expenses by trimming unnecessary information from the context window. Its Foundry IQ feature provides a shared knowledge layer that uses subqueries and reranking to return specific grounded passages rather than entire documents. Microsoft reports that this approach has improved evidence recall by up to 54% while reducing retrieval-token costs by 34%. Additionally, its tool search functionality aims to reduce input-token consumption by allowing models to describe needed capabilities rather than receiving complete tool catalogs.

Google’s Agent Development Kit (ADK) addresses context management by offering four distinct storage locations for TypeScript applications: session events, session state, long-term memory, and artifacts. This structure is designed to prevent bugs and inefficiencies that occur when all information is placed solely within the prompt, such as stale data growth, excessive costs, or the loss of user preferences between sessions.