wtf is pi-supermemory?

pi-supermemory adds persistent memory to Pi, a coding agent. Normally an AI coding assistant forgets everything once your session ends. This package changes that by connecting Pi to a memory system called Supermemory, so it can recall things it learned in earlier sessions, like which test command your project uses or a decision you made about how a feature should work. By default the memory database lives on your own computer through a local server called Supermemory Local, so nothing is sent to an outside service unless you choose to use hosted Supermemory instead. Before Pi responds to you, it quietly checks this memory for anything relevant and adds it to its context. After a useful conversation, it automatically saves durable facts such as preferences, bug fixes, and setup notes, based on keywords like remember, decision, bug, and fix appearing in the conversation. You can also manage memory directly with commands inside Pi, such as saving a note, searching past memories, checking your saved profile, or telling it to forget something. Plain language works too: you can simply say things like remember that this repo uses Bun, and it will store that. Memory is split into two separate scopes: one tied to you as a user, and one tied to the specific project you are working in, so personal preferences do not leak between unrelated codebases. There is also a privacy feature that redacts anything wrapped in private tags before it gets stored, so secrets like API keys typed into a conversation are not saved in plain text. Setup involves installing the Pi package, then running the separate Supermemory Local server, which itself needs an OpenAI key or a local model to do its own summarizing work behind the scenes. The project is released under the MIT license.