Where should an emotional-continuity state layer live in an agent framework?

[pioneerjeff]

Hi, I’m building Emotion Engine, a small local-first emotional-continuity state layer for long-running agents.

The current v0.2.2 release adds a local stdio MCP server plus client registration docs, but I’m still trying to understand the best integration boundary for agent frameworks.

The core idea is: the LLM or host runtime decides what happened; Emotion Engine only persists a compact state packet such as slow mood, short-lived affective pulse, trust movement, and boundary guidance.

From your framework’s perspective, would this fit better as:

1. a sidecar state packet,
2. an MCP/tool plugin,
3. a normal memory summary,
   or not as part of the framework at all?

Repo: https://github.com/pioneerjeff-labs/emotion-engine
Demo: https://pioneerjeff-labs.github.io/emotion-engine/demo/
MCP docs: https://github.com/pioneerjeff-labs/emotion-engine/blob/main/docs/MCP.md