*Proposal Doc: https://docs.google.com/document/d/18pB_JkfDdxvRh0Zl1e9LE8ny1m-2_E8A1D0Dtkm3cg4/edit?tab=t.0#heading=h.w9x5jty6sxmd*

Context

Google ADK has become one of the fastest-growing agent frameworks, but today only relational and Google-managed databases have native persistence support. MongoDB is currently limited to MCP integrations, which solve a different problem than the agent infrastructure itself. This proposal explores adding MongoDB Atlas as a first-class backend for ADK’s SessionService and MemoryService, leveraging an existing prototype and validating whether Google will support a native or community integration path.

Definition of done

• Validate with Google ADK maintainers whether native or community integrations are feasible.
• Estimate the engineering effort to implement Atlas-backed SessionService and MemoryService for the Python SDK.
• Produce a recommended implementation approach (native vs. community integration), including scope, dependencies, and documentation requirements.

Pitfalls

• Google may not accept third-party database implementations into the core ADK project, even for Google products (e.g. Firestore).
• Native integration may not be achievable, requiring a community integration with lower visibility.