Start here

MAGLA Memory is a small cloud API for carrying verified work across agent sessions — not a chat-history dump and not another database you have to run.

Start with the proof, not the architecture

If you have five minutes, do the two-session quickstart. You will save one result in one process and recover it in another.

The public surface is deliberately small

statusIs this project and memory service ready?preflightWhere should this task resume?contextWhich verified checkpoints matter now?record_outcome / record_failureClose the loop with evidence-backed state.

What MAGLA Memory does not require

No Docker or self-hosted memory server.
No vector database or RAG pipeline.
No commitment to one model vendor.
No credit card for the free developer project.