Feature Request: Cloud-Based Session Storage for Session Management

[tahitimoon]

I really like this library and am planning to use it in a production environment. However, the current session management only supports storing sessions on the local file system, which is not very convenient to manage. For building production-grade systems, cloud storage is essentially a requirement.

It would be great if session storage could be extended to support cloud storage backends such as S3 buckets. I recently noticed that Strands Agents already provides this capability, and I’m wondering whether the Claude Agent SDK has plans to add similar support.

Reference documentation:
https://strandsagents.com/latest/documentation/docs/user-guide/concepts/agents/session-management/

[cdnsteve]

What about local session management instead of dependency of a third-party like S3? Would that fit the need @tahitimoon ?

[tahitimoon]

What about local session management instead of dependency of a third-party like S3? Would that fit the need @tahitimoon ?

I think there might be a slight misunderstanding - the current implementation already supports local session management (file system storage), which works great for development and single-server deployments.

My feature request is specifically about adding cloud storage options for production environments where:

1. Horizontal scaling: When running multiple server instances behind a load balancer, sessions need to be accessible across all instances
2. Reliability: Cloud storage provides built-in redundancy and persistence that local file systems may not guarantee
3. Deployment flexibility: Many modern deployment platforms (containerized environments, serverless, etc.) have ephemeral file systems

So ideally, the SDK would support both approaches:

• Local file system (already exists) - perfect for development and simple deployments
• Cloud storage backends (S3, Azure Blob, GCS, etc.) - for distributed production systems

The goal isn't to replace local storage, but to provide additional options for different use cases. Think of it like how many database ORMs support both SQLite (local) and PostgreSQL/MySQL (remote) - different tools for different scenarios.