Every new chat starts from zero. MemorySafe connects your agents’ memory into one — a single file on your computer, recalled only when it matters, and never uploaded anywhere by MemorySafe.
No account. No email signup. Code on GitHub
You: Remember: our launch moved to Friday.
✓ Saved to MemorySafe.
MemorySafeOne shared memoryLaunch → moved to Friday
Available to your connected assistants
You: What should I finish first?
↳ Retrieved from MemorySafeThe launch is Friday, so start with the signup page.
Different assistant. Same context.
Every chat starts at zero. You re-type your tools, your rules, your deadlines — every session, in every app.
Claude knows some of it, ChatGPT another slice, Codex nothing. Your context is fragmented across vendors that don’t talk to each other.
Vendor memory lives on their servers, on their terms. You can’t open it, back it up, or take it with you.
Deliberately boring infrastructure: capture facts, keep them in a file you own, recall them when they matter.
As you work, MemorySafe saves concise, durable facts — decisions, preferences, deadlines. Automatically if you turn that on, or only when you say “remember this.”
Everything lands in one database file on your computer. Open it, back it up, review it in the local dashboard.
Your assistant searches memory before it answers, and only the facts it needs enter the conversation. Memories are never pruned for being old.
Models are frozen the day they ship. Your memory is the part that keeps learning — every decision and preference you confirm compounds.
Your name, your tools, your rules of engagement. No more introductions.
What you chose, what you set aside, and why — settled questions stay settled.
Your standards, your edge cases. Answers start from your context, not from zero.
Start on your computer today. A shared memory across devices is what comes next.
A shared place for your supported desktop assistants to save and recall useful context.
One shared memory across supported AI apps and devices — ChatGPT, Claude, Codex and more — designed to stay current wherever you work.
The free local list is what works today. The hosted edition is where every assistant you use — ChatGPT included — meets one memory.
Most privacy pages ask you to trust a policy. Ours asks you to look at a file path.
Read the code on GitHubYour memory database stays on your computer. No account, no sync, no MemorySafe server behind the free plugin.
Protect locks a memory against change. Forget removes it from recall. The local dashboard shows what’s saved and what changed.
Secrets, credentials, payment and government IDs, health data: never captured, in any mode.
A connected AI service still processes the context you choose to share, and forgetting removes a memory from recall — not from history or backups. We say both plainly.
The free plugin keeps its database on the computer where you install it. It can be shared by supported assistants on that computer. Cross-device memory is planned for the hosted edition, which is still in development.
The memory database stays on your computer. A connected AI service can still process the context you choose to share under its own terms. Local storage doesn’t mean the AI service itself runs locally.
Use the local dashboard to review saved memories, inspect changes, manage protection, and remove memories from recall. Removing a memory from recall does not completely erase history or backups.
We’re testing tools to prioritize memories automatically: what to keep, revisit, or set aside. They’re planned for the paid hosted edition and aren’t included in the free local beta.
