The notebook
Your AI knows everything. Except you.
MemoryPlugin changes that. It learns who you are and how you work, then remembers it across every AI tool you use: ChatGPT, Claude, Gemini, and 18 more.
30-second setup · 7-day free trial · The depth is there when you want it
Prefers TypeScript, strict mode, no default exports
Work
Runs a weekly farmers market stall on Saturdays
Personal
Rebrand launch is October 12, tone: warm, direct
Acme project
Saved from a Claude chat. ChatGPT knows it from your next conversation.
The problem with generic
Generic answers aren’t a model problem. They’re a memory problem.
Imagine a doctor who gives every patient the same prescription. You’d walk out.
But that’s the AI you’re using. The same advice for you, your competitor, and a million strangers. Not because it isn’t smart. Because it doesn’t know you.
So you compensate. You re-introduce yourself. You paste the same context. You answer the same questions you answered last week, in another tab, to the same machine.
The smartest AI in the world, and it treats you like a walk-in.
With and without
Same question. Different AI.
Nothing changed but the memory.
Without memory
Can you help me plan next week?
Sure! A good weekly plan starts with your priorities. Try listing your most important tasks, blocking time for deep work, and leaving buffer for the unexpected. Would you like a general template?
Helpful for anyone. Useful to no one in particular.
With MemoryPlugin
Can you help me plan next week?
You ship Fridays, so keep Thursday for the Acme QA pass. The rebrand copy has to lock this week to hold the October 12 launch, warm and direct like you wanted. And nothing on Saturday: market stall.
Same model. Same prompt. The memory did that.
Getting to know you
It learns you as you go.
No forms, no setup project. Memory builds itself while you work.
Answer a few questions
A conversational get-to-know-you flow asks about you and your work, and you watch memories appear live as you answer.
Captured as you chat
As you talk to any connected AI, details worth keeping get saved mid-conversation. No buttons, no copy-paste.
Nothing goes stale in a tab
Reopen an old chat and the extension shows what changed since you were last there: 2 new, 1 updated. One click brings just those in.
…we’re a two-person studio, all our client work is Next.js, and we ship every Friday.
2 new · 1 updated since you were here
It compounds. Week one, it knows your stack. Month three, it knows how you think.
The part everyone gets wrong
A memory that stays clean.
You’ve seen what happens to built-in AI memory. It fills up. It hoards junk. It confidently repeats facts that stopped being true a year ago. MemoryPlugin was built around not letting that happen.
Buckets
Work, Personal, each client: separate buckets, zero bleed. The AI only sees the bucket you’re in, so contexts never mix and nothing feels cluttered. Unlimited, on every plan.
Memory suggestions
A background pass finds duplicates, contradictions, and stale facts, then proposes the fix: merge these two, update that one. You approve, bucket by bucket. The mess never accumulates.
And it stays light: Smart Memory sends summaries up front and full detail on demand, so memory costs about 90% fewer tokens in your context window.
Recall
Ask for anything. It’s already there.
Roadmap whiteboard, March planning session
Image memory · Work
Q2 priorities: memory quality first, then the desktop app
Memory · Work
Search by meaning
Hybrid search finds memories by exact keyword and by what you meant, so “that client with the strict brand rules” works as well as the client’s name.
Ask your memories
Plain-language questions over everything you’ve stored, with real answers drawn from your own context, not the internet’s.
Images too
Save images as memories and find them later by describing them. The whiteboard photo from March is one sentence away.
Everywhere, and yours
One memory across ChatGPT, Claude, Gemini, and 21+ AI tools.
MemoryPlugin doesn’t belong to any one AI company. It sits outside all of them, which is exactly why it can serve all of them.
Browser extension
Chrome and Safari, desktop and mobile. Works right inside ChatGPT, Claude, Gemini, and the rest of your tabs.
MCP server
Hosted or local. Claude Desktop, Claude Code, Cursor, Windsurf, and 100+ MCP clients connect by URL.
Custom GPT, TypingMind, API
A Custom GPT for ChatGPT’s apps, a native TypingMind plugin, and a documented API for everything else. See all platforms.
One-click export
CSV, JSON, or Text with AI-ready instructions prepended. Your memory is yours to take anywhere. No lock-in.
Connected
- chrome · extension ✓
- safari · extension ✓
- mcp · memoryplugin ✓ connected
- ChatGPT → MemoryPlugin GPT ✓
- typingmind · native plugin ✓
- POST /api/v2/memory → 200
export → csv · json · text
For the power users
As deep as you want to go.
None of this is homework. It’s range.
Acme projectbucket
Acme rebrand notesmemory · v3
v3 · edited in May · view diff · restore
Keyboard-driven. Command-palette search, an instant bucket switcher, shortcuts for everything you do twice.
Version history. Every edit is versioned. Open a memory’s history, see exactly what changed, restore any earlier version.
Bulk add. Paste a list, one memory per line, and duplicates get caught on the way in.
Files and presets. Upload PDFs and docs for Q&A, save images as memories, start buckets from category presets.
Two years in. Built with the people who use it.
May 2024
Shipping since
Monthly
Releases, no gaps
3,800+
Users
4.5★
From 100+ reviews
Memory is a long game, and MemoryPlugin is built like one: independent, funded by the people who use it, shipping every month since May 2024. The roadmap is built in the open. Bulk add, export formats, in-chat detection: all of it exists because users asked.
Independent · Funded by its users · Public status page · Export anytime
Objections, welcome
Questions you’re probably asking.
What happens when ChatGPT’s memory is full?
ChatGPT’s built-in memory has a cap, and when it hits, saving quietly stops. MemoryPlugin has no memory cap on any plan. Store as much as you want, keep it organized in buckets, and the suggestions system keeps it from turning into a junk drawer.
How is MemoryPlugin different from built-in memory?
Built-in memory lives and dies inside one product. MemoryPlugin is one memory that every AI you use shares: tell Claude something once and ChatGPT knows it from your next conversation. It’s organized in buckets, curated by suggestions, and exportable in one click.
Does it work with Claude, Gemini, and my other tools?
Yes. The browser extension covers ChatGPT, Claude, Gemini, Grok, and the rest of your AI tabs. The MCP server connects Claude Desktop, Claude Code, Cursor, Windsurf, and 100+ other MCP clients. There’s also a Custom GPT, a native TypingMind plugin, and an open API. 21+ platforms in all.
Will the AI mix up my work and personal contexts?
No. Buckets keep contexts fully separate, and the AI only sees the bucket you’re working in. Work stays work, clients stay separate, personal stays personal.
Won’t a big memory eat my context window?
Smart Memory sends compact summaries up front and pulls full detail only when asked, which cuts memory’s context footprint by about 90%. A two-year-old memory costs barely more than a two-week-old one.
Can I take my memories with me?
Always. One-click export to CSV, JSON, or Text, and the Text format ships with AI-ready instructions so you can paste it anywhere. No lock-in, ever.
What happens to my memories if I cancel?
Cancel in the dashboard whenever you like, and export everything first in one click: CSV, JSON, or Text, with the Text format carrying AI-ready instructions. Your memory keeps working wherever you go next.
How much does it cost?
Every plan starts with a 7-day free trial. Core is $79 per year and includes unlimited memories across all 21+ platforms. Pro adds your entire chat history and the intelligence layer on top.
Comparing options? See how MemoryPlugin stacks up against built-in memory.
Make it yours.
Set up in 30 seconds. Free for 7 days. Your AI will thank you.
Start free trialMemory is half the story. The other half is everything you’ve already discussed: see Chat History.