How to Transfer Your Gemini Memory to OpenClaw

How Gemini stores memory

Gemini's memory is less structured than ChatGPT's or Claude's. Rather than automatically saving preferences from conversations, Gemini relies mainly on a manual "Instructions" system: things you explicitly tell it to remember through Settings and Help > Instructions for Gemini.

If you haven't touched those settings, you might not have much stored. That's okay. You can write your preferences from scratch and use this guide to seed OpenClaw properly.

How OpenClaw handles memory

OpenClaw stores memory as plain text files on your computer. The key file is MEMORY.md, which lives in your workspace folder and gets loaded into every conversation with your agent.

It's simple and transparent. The catch is that recall isn't guaranteed. OpenClaw does its best to reference what's in the file, but it makes its own judgment call about when that information is relevant.

Why trimming matters

Even if Gemini doesn't hand you back a huge export, you still want to be deliberate about what goes into MEMORY.md. OpenClaw reads that file on every single message you send. Every word in there costs you tokens on your LLM API, even if most of it isn't relevant to what you're asking right now.

People who've tried pasting in everything they can think of usually come to the same conclusion: treat MEMORY.md like a business card, not a biography. Name, location, a few key projects, core preferences. The stuff that's almost always relevant no matter what you're talking about. Everything else is just noise that slows down responses and runs up your API bill.

The sweet spot is around 20-30 entries. If Gemini only gave you 15 good ones, that's perfectly fine. Less is more here.

Step 1: Export your Gemini memory

Start by checking what Gemini has stored. Go to Settings and Help > Instructions for Gemini.

Then open a Gemini conversation and paste this prompt:

If Gemini gives a short response, follow up with: "Are there any other preferences or details you've picked up from our past conversations?"

If Gemini says it doesn't have any saved info, write your preferences by hand. A few lines covering your role, projects, and how you like responses is enough.

Step 2: Trim it in the ImportMemory editor

Paste whatever Gemini gave you into the ImportMemory editor. Cut things like:

• Anything no longer accurate: old jobs, finished projects
• Vague inferences ("you seem interested in productivity")
• Duplicate entries
• Anything too personal for an API

Keep: your current role, response preferences, key tools, and standing instructions. Aim for 20-30 entries.

Step 3: Paste it into MEMORY.md

Open ~/.openclaw/workspace/MEMORY.md and paste your trimmed memory with clear headers:

## Identity
- Name: [your name]
- Location: [city, country]
- Role: [job title]

## Active Projects
- [Project name]: [brief description]

## Preferences
- Keep responses concise
- Use bullet points for lists
- Skip the disclaimers

Save and restart with openclaw gateway.

Tips for a better result

• Gemini exports are often thin. If you didn't get much back, write your core preferences by hand. Five to ten well-written lines beats a messy export.
• Keep it lean. 20-30 focused entries. More than that and you're paying tokens for stuff that rarely comes up.
• Update it regularly. MEMORY.md is just a text file. Review it every month or two.
• Use clear formatting. Grouping by Identity, Projects, and Preferences helps the model find what it needs.
• Recall isn't perfect. If something important keeps getting missed, state it directly in conversation.

Frequently asked questions