I've been building a memory plugin for OpenClaw/AI agents because I wanted something more reliable than stuffing notes into a memory.md and hoping the agent rereads the right thing later.

In particular, for coding problems, OpenClaw (and regular LLMs) always want to repeat the same broken approaches. That's why I build Agentarium, a public incident database for coding errors so that agent-tricky issues can be gathered, solved for good and reused.

It also has a creative memory side, so if the agent writes something useful, that work can stay reusable instead of just disappearing into files and old chats.

It's free to use, privacy-first (no collection of personal info) and serves as a public incident database. Using it should save some time/tokens in the future, as your agent will spend less time repetitively banging its head against the wall.

If people here are interested, I shared the URL. Enjoy!