MongoDB and Google ADK connect: agent sessions and memory no longer need a separate store
Synopsis
MongoDB released an integration with Google ADK that gives agent state durable memory. Previously developers still had to build their own tooling for a data store that could retain agent context over time. The integration uses MongoDbSessionService to persist sessions, events, and application-, user-, and session-scoped state to MongoDB, and MongoDBMemoryService to ingest session events into collections and recall memories through keyword matching; retrieval reuses MongoDB vector search combined with full-text search. It supports MongoDB Atlas and self-managed MongoDB 8.0+ deployments.
Interpretation
An agent's session state and long-term memory can now live in the same MongoDB as its application data, instead of separate systems for retrieval, state, and memory. Previously developers still needed to build their own tooling for a data store that provides durable agent memory for agent state. The integration persists sessions, events, and application-, user-, and session-scoped state to MongoDB, and keeps conversation memory independent of any single machine so a user's history carries across sessions.
The session service holds the state multi-turn interactions need, and the memory service turns session events into memories that can be recalled. MongoDbSessionService includes optimistic concurrency control to manage concurrent updates safely, and configurable collections make it easier to fit existing data architectures. MongoDBMemoryService ingests session events into MongoDB collections and recalls relevant memories through keyword matching; memory ingestion is idempotent, so adding a session again does not create duplicates.
Retrieval is handled by MongoDB's built-in search: vector search uses Voyage AI embedding models, and the $rankFusion aggregation stage can combine vector search with full-text search into hybrid search. The application owns the database connection, database name, and search configuration, while the agent gets focused tools to call without exposing those operational details in its context. The integration supports vector and hybrid search on MongoDB Atlas and self-managed MongoDB 8.0+ deployments, with index names, embedding fields, limits, and candidate counts configured through MongoDbToolSettings.
Perspective
The integration targets developers building agents with Google ADK, providing session and memory storage plus MongoDB-based vector and hybrid retrieval, and supports MongoDB Atlas and self-managed MongoDB 8.0+ deployments.
The source gives no performance, latency, or reliability measurements for the integration and does not report behavior under real production load; memory recall uses keyword matching, while semantic retrieval is backed by Voyage AI embedding models.