Hi @modelscope maintainers,
I’m Vivek Gupta, Founder & CEO at Memcode. For applications built with ms-agent, a user may want stable, explicitly approved project context to remain available across separate agent runs, even when the orchestration itself is stateless or distributed.
Would a small MCP integration example be useful? It would keep long-term personal context distinct from task traces and model prompts, require user consent for saves, and use OAuth/DCR with PKCE for the hosted connection. No full conversation or secret would be sent by default, and the feature would remain optional.
Which ms-agent integration point would you want a first PR to target?
Hi @modelscope maintainers,
I’m Vivek Gupta, Founder & CEO at Memcode. For applications built with ms-agent, a user may want stable, explicitly approved project context to remain available across separate agent runs, even when the orchestration itself is stateless or distributed.
Would a small MCP integration example be useful? It would keep long-term personal context distinct from task traces and model prompts, require user consent for saves, and use OAuth/DCR with PKCE for the hosted connection. No full conversation or secret would be sent by default, and the feature would remain optional.
Which ms-agent integration point would you want a first PR to target?