I have been working to add a memory system to TORCH. So every project it's deployed into, agents can generate and save memory artifacts for that repo.
It working in a similar way to how OpenClaw's memory management works but on a per-repo basis.
You could hook TORCH memory up to your AGENTS.md or CLAUDE.md to enable every agent working on the project to utilize and append to the repo memory.
TORCH: AI Workforce Orchestration