I’ve been following Cole Medin’s Archon, and I’m seriously impressed by its ability to build other AI agents. However, I came across something that could take Archon to the next level—Persistent Memory.
Recently, another dev, VRSEN (Arseny), released a feature for his agent-building AI (agency-swarm) that allows it to install and retain memory. This means no more AI going off the rails just because it lost track of its task. It stays focused, remembers previous steps, and functions much more effectively.
Imagine Archon with true memory persistence—it could retain progress between tasks, refine its own workflows over time, and become dramatically more powerful.
Cole, would you consider integrating a memory feature like this into Archon? I truly believe this is a must-have for AI agents, and Archon would benefit massively from it.
@ColeMedin can correct me if I’m saying something daft here. However, my understanding is that the current version of Archon manages agent workflows with LangGraph which has sophisticated memory management features:
I appreciate the suggestion a lot @PilotHobs, and please, feel free to make a feature request for it on GitHub! Arseny has a great YouTube channel and I did actually see this video.
@Groove Yes LangGraph has amazing memory management features, but it’s all managing the memory within a specific conversation/workflow execution. I believe what @PilotHobs is getting at is more persistent memory across uses of Archon so it can improve over time based on previous agents it built for the user.
I agree this is super important for Archon long term! I’d love to integrate with something like Mem0 for this.
Check out the memory feature here: Agency-Swarm Memory
How VRSEN Built Persistent Memory for AI Agents