yigraf keeps one connected graph over your code, what it's for, what's left, and why it ended up this way โ and hands your agent exactly the right slice, right when it acts. It stops starting over.
Every /clear wipes what it knew. And the code never told the whole story โ it records what runs, never why it's shaped that way or what it must guarantee. So every session your agent relearns the repo, re-reads files it already understood, and sometimes undoes decisions you settled weeks ago. yigraf remembers, so it doesn't.
The ones an agent loses on every reset โ kept as a living graph beside your code, with the links between them doing the real work.
A task points at the code that implements it. A decision is pinned to the code it concerns. So when the code drifts away from what was supposed to hold, yigraf notices โ ask it what governs this file? and it can answer.
It listens on your repo and speaks up at the right moments โ but you can always prompt it in plain language.
Your agent asks yigraf for the governing slice over MCP. Works with every MCP-capable host โ the universal floor.
yigraf injects the slice the moment your agent edits a file, at the best fidelity your host's hooks allow.