AI coding agents are getting better, but they still start most sessions with amnesia.

I use Codex, Claude Code, Cursor, and other agent workflows every day. The annoying part is not code generation. It is context repetition: what the project is, what I tried yesterday, why a decision was made, which logs mattered, and what the next agent should not forget.

That is why I built KnowYou MyWiki.

KnowYou MyWiki is a local-first Mac app that turns existing messages, logs, notes, and work traces into a source-linked personal wiki for AI agents. The goal is to make your working memory inspectable and reusable by future agents, without making your private work context a cloud profile by default.

What the first version does:

• collects local working traces
• turns messages and logs into readable memory
• keeps source evidence attached
• creates a MyWiki-style personal knowledge layer
• exports memory for agent workflows

I care about the local-first part because working context is sensitive. It can contain code decisions, debugging notes, product strategy, private links, and half-formed thoughts. I want users to own it, inspect it, and decide which agents can use it.

Demo: https://www.youtube.com/watch?v=UiaAVBtqBx0
Product: https://giiift.site/know-you/

I would love feedback from developers using AI coding agents:

What context do you repeat most often to agents?
Would you trust a local-first memory layer more than a cloud memory product?