Per Client Memory Router

Stops one agent working N clients from bleeding context between them, and stops N parallel client jobs from starving each other on the same model quota. Partition memory into a global store plus one isolated store per client, then gate execution through a concurrency-limited router so no single client's job queue can starve the rest. Use when one agentic setup serves multiple clients, when a client's context/preferences ever showed up in another client's output, or when parallel client jobs are timing out or burning quota unevenly.

search-atlas-group Updated 0 repo stars

File contents

search-atlas-group/amm-founding-circle commit aac4932a86

Frequently asked questions

npx get-skill add search-atlas-group/per-client-memory-router