Tired of re-explaining yourself to every AI tool?

Posted by ZestycloseTie1793@reddit | LocalLLaMA | View on Reddit | 7 comments

I use multiple AI agents for different things — OpenClaw for general tasks, OpenCode for coding, sometimes Hermes for quick stuff. Every single one forgets who I am between sessions.


Tried the built-in memory features. Problem is they're locked to one tool. OpenClaw's memories don't transfer to anything else. Each agent is an island.


So I made Relic — a set of Markdown files that any AI agent can read to learn about you and your preferences. It's based on the Relic biochip from Cyberpunk 2077 (the thing that stores Johnny Silverhand's soul).


Your AI gets:
- A personality file (SOUL.md) — so it knows who it is
- A user profile (USER.md) — so it knows who you are
- A memory file (MEMORY.md) — so it remembers what happened
- An agent registry — tracks which agents have connected


All plain text. `cat` readable. No database, no server, no install step other than `git clone`.


The cross-agent memory sync is the killer feature for me. I can work with OpenCode all morning, it writes memories to the shared file, then when I switch to OpenClaw in the afternoon it picks up where OpenCode left off. Like one consciousness jumping between bodies.


GitHub: https://github.com/LucioLiu/relic


Anyone else dealing with AI memory loss across tools? How are you handling it?