Shared memory
Each agent keeps its own memory, so what you tell Claude, ChatGPT doesn't know. Keep one memory in Ether instead: every agent you connect recalls it at the start of a task, and adds to it as you work.
Set it up
Paste into any agent that supports skills. It installs the Ether memory skill, which tells the agent how to keep its memory in your workspace.
Set up Ether memory for me. Download https://etherfs.app/skills/ether-memory/SKILL.md and save it unchanged as ether-memory/SKILL.md in the folder you load my personal skills from. If you can't install skills, tell me instead of saving it anywhere else. Then check that you can reach my Ether workspace, and if your own memory feature holds anything about me, offer to copy it into memory/.
How it works
Each memory is a file holding one fact: who you are, how you like things done, or a decision a project made. They're kept in a folder per project, with an index that the agent reads first.
memory/
launch/
MEMORY.md
code-comments.md
pricing-decided.md
website/
MEMORY.md
copy-style.md A memory starts with a line that says what it's about, so an agent opens only the ones that bear on its task.
---
name: code-comments
description: Comment only what the code can't say (the why)
metadata:
type: feedback
---
Skip comments that restate a name, a type or the next line.
**Why:** they go stale and the authority is the code. Your memories are yours
They're files in your workspace, so you can read, change or delete any of them in the web app, and get one back from its history. To carry a conversation's work rather than a fact, hand it off.