Activegraph

activegraph is an open source Python runtime for long running AI agents from Yohei Nakajima, the creator of BabyAGI. It records everything an agent does in an append only event log and projects that log into a live graph of beliefs, tasks, evidence, decisions and dependencies. The graph becomes the agent's world: what exists, what depends on what, what was produced, what was approved and why it changed. It is aimed at engineers and researchers building agents that run for a long time and need to be inspected and reproduced. Because state is derived from the event log, any run can be replayed, forked at a specific event, or diffed against another run, and the origin of every object can be explained. Forks replay the shared prefix from cache, so earlier LLM calls are not paid for twice. A distinctive primitive, relation behavior, places coordination logic on the edges of the graph rather than duplicating it across nodes. The documentation is written to be read by agents as well as people, with llms.txt files and a quickstart that a coding assistant can follow to build a working experiment. Reference packs and a lab repository provide composable starting points. activegraph is free and licensed under Apache 2.0, requires Python 3.11 or later, and installs with pip. No commercial plan is published. The project presents itself as the architectural successor to BabyAGI and ships frequent releases, with a paper and a conference talk explaining its design.