wtf is infinicon?

Infinicon is a memory system for AI agents, built as a software development kit plus a small server that you run yourself. The idea is that AI agents normally forget everything once a conversation ends, and Infinicon gives them a place to store and retrieve information across sessions, so an agent can remember something like a user's preference and bring it back up later. The project has two main pieces. The sdk, called @infinicon/sdk, is what you install in your own application to talk to the memory system, using simple calls like remember and recall. The server, called @infinicon/server, is what actually stores the data and answers requests. You can run it locally with a single command, or deploy it to Render. By default the server keeps everything in memory, meaning data is lost if the server restarts, but you can connect a Postgres database instead if you want it to persist. Infinicon is written in TypeScript and needs Bun to run the server. It exposes a set of API endpoints for tasks such as health checks, ingesting new memories, querying for relevant ones, and consolidating stored information. There is also a plugin host, which lets other developers add their own storage or ranking logic instead of relying only on the built in lexical, or keyword based, search. The repository includes example code to help newcomers get started, including a one file chat agent example and a fuller example that shows memory recall and consolidation together. Documentation covers the project's vision, a glossary of terms, its architecture, and a roadmap, plus formal specifications and decision records for those who want to understand the reasoning behind design choices. This project is best suited for developers building their own AI agents or chatbots who want those agents to retain context between sessions rather than starting fresh every time. It is released under the MIT license, so it can be used freely, including for commercial projects, as long as the copyright notice is kept.