Could ACI expose durable memory at agent workflow boundaries?

Hi, I’m Vivek Gupta, Founder & CEO at [MemCode](https://memcode.in).

ACI.dev gives agents a broad tool surface through direct function calling and MCP, where continuity can make repeated workflows much more useful. A durable memory layer could retain user or workspace preferences, successful tool sequences, prior workflow context, and explicit decisions across separate agent sessions without coupling memory to any one model.

Would you be open to an optional MemCode integration at the agent/session and tool-execution hooks: retrieve scoped context before a workflow and write only explicit, approved outcomes afterward? It can preserve provenance, retention, and deletion controls and remain a clean no-op for users who do not configure it. I’d be glad to submit a focused adapter/example PR or discuss the smallest integration point on a short technical call.

The issue names no files or tests. Start by reviewing the agent/session and tool-execution hooks used by direct function calling and MCP, then define the smallest optional integration boundary. Done would include scoped context retrieval, explicit approved outcome writes, provenance, retention and deletion controls, with a no-op when unconfigured.

Written by the indexing model from the issue text.