Where Is Your Assistant's Brain? Understand Memory and Workspaces

Ever had that painful moment where you spent half a conversation teaching an AI something, only for it to forget everything next time?

This article is here to solve that frustration.

What You'll Learn

• Understand what a workspace is and why it matters
• Learn how memory works
• Know where your data is stored
• Back up all your settings with one command

What Is a Workspace?

Think of a workspace as your assistant's private office:

• Files it reads, documents it writes, and scripts it runs all happen inside that folder
• You can create multiple workspaces so different jobs stay separated

~/.openclaw/workspace-main        ← office for your main assistant
~/.openclaw/workspace-project-a   ← office for project A
~/.openclaw/workspace-project-b   ← office for project B

That way, "organize my personal journal" and "write code for project A" do not get mixed together.

How Does Memory Work?

OpenClaw stores important pieces of conversation as memory, so the assistant can search and reuse them later.

In plain English: your assistant remembers things you told it before.

⚠️ If it feels forgetful, check these:

• Is memory enabled? Semantic search may require an OpenAI API key
• Was old memory cleaned up, compacted, or limited by length?
• Did you start a brand-new session? New sessions start with an empty short-term context

Where Is the Data Stored?

One-Command Backup

Your entire OpenClaw life mostly lives under ~/.openclaw. Back it up regularly:

# Copy everything to a backup location
cp -r ~/.openclaw ~/backup/openclaw-backup-$(date +%Y%m%d)

💡 Moving to a new computer? Copy the whole ~/.openclaw folder and you can often continue without running onboarding again.

Next Up

Workspace basics are done. Next we move into the fun part: multiple assistants with different roles, routed by groups, bots, or use cases.

← Previous: Skills | 👉 Run multiple assistants behind one Gateway with Bindings