[Feature request]: I'd like to save specs in AI memory framework instead of Markdown

[clay-good]

Note

Moved from issue #745, originally opened by @qdrddr on 2026-02-22. Continuing the conversation here in Discussions — the original issue thread (including any comments) stays available at #745 for the record.

---

Problem or motivation

AI agents generate MD files a lot.
Explanation, summary, overview, plan, design, diagram, you name it.

And they forget about them, occasionally bump into one, and start to update it like crazy. Create duplicates. Partial updates. Or find some really old one no longer relevant and start making changes to your code.

While keep burning my tokens.

Proposed solution

Implement a graph memory that keeps connected specifications linked and reduces duplicates. So when the update happens it updates relevant parts of the text.

Additional context

I propose to implement a graph-based memory to store the specification:
https://github.com/vectorize-io/hindsight