Remember: Draft the launch post · --task --high --assign @writer --notify @editor
launch post · assignedLaunch post: done ✓ by @writer, reviewed by @editor
Tell Claude something in the morning. ChatGPT remembers it at lunch. JauMemory gives every AI you use one shared memory, so you never re-explain yourself again.
Private beta · invite-only
Not an SDK you build into your app · a shared memory across the AI tools you already use. No code.
Under the hood: 50 tools · agent teams · credential vault · REST APISimulated right here in your browser · no account needed
Example: you tell Claude "Remember: my daughter's recital is Friday at 6pm" and it saves the memory. Later you ask ChatGPT "What's happening Friday?" and it answers from that same memory.
Example: a project-manager agent saves the task "Draft the launch post", assigned to a writer agent with the editor notified. The writer gets the notification, drafts the post, and saves it to the shared memory. The editor gets notified, reads the draft from memory, and marks it reviewed. The project manager sees the completed work, all through one shared memory.
Every agent reads and writes the same memory · assignments and notifications built in. See agent tools →
Why JauMemory
Your projects, your preferences, the way you like answers: say it once and every new chat already knows.
No more pasting the same background into every conversation.
Hit a rate limit, or just like a different app for different jobs? Your memory comes with you.
Start planning a trip in ChatGPT over lunch, finish it in Claude tonight.
Every conversation adds to a history your AI can draw on, so answers keep getting more personal.
A month in, it remembers your team, your goals, and that recital on Friday.
How it works
Three steps. No files, no code.
JauMemory is in private beta. Request an invite, and once you're approved, create your account in a couple of clicks.
In Claude or ChatGPT, open Settings, choose Connectors, and add JauMemory. That's the whole setup.
Tell one AI to remember something. Ask any of your other AIs about it later. That's it.
Then open any other tool and recall it. That's the point: one memory, everywhere you work.
Prefer a config file? Desktop setup →
What you get
One login covers memory, search, agents, a credential vault, skills, and scheduling.
Remember in Claude, recall in ChatGPT. One memory shared by every AI you use.
Encrypted at rest with keys unique to you, encrypted in transit, and walled off from every other account.
Ask in your own words and JauMemory finds the memory you meant. Your content is never sent to a third-party AI API.
Your memories file themselves: wins, fixes, ideas, and questions get sorted, and duplicates get merged into insights.
Run more than one AI assistant? Give each an identity and a shared workspace so they build on each other's work.
Store API keys once, encrypted. Your AI agents can use them without ever seeing them.
Teach your AI routines, then run them on a schedule. Memory that doesn't just sit there.
Add JauMemory from your AI app's connector settings, approve in your browser, done. No passwords pasted into a chat.
Every user's memories are encrypted with their own key.
All traffic between your tools and JauMemory is encrypted.
Database-enforced separation between accounts.
Device-bound, revocable logins. Nothing secret pasted into a chat.
Your data is yours. Export it whenever you want.
Under the hood
The same memory, programmable: an MCP server with 50 tools, a REST API, and full docs. Per-client config below.
# Add to claude_desktop_config.json
# macOS: ~/Library/Application Support/Claude/
# Windows: %APPDATA%\Claude\
{
"mcpServers": {
"jaumemory": {
"command": "npx",
"args": ["-y", "@jaumemory/mcp-server"]
}
}
}
# Then use mcp_login to authenticate
# Install via CLI
claude mcp add jaumemory -- npx -y @jaumemory/mcp-server
# Or edit ~/.claude.json
{
"mcpServers": {
"jaumemory": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@jaumemory/mcp-server"]
}
}
}
# Then use mcp_login to authenticate
# Create .cursor/mcp.json
{
"mcpServers": {
"jaumemory": {
"command": "npx",
"args": ["-y", "@jaumemory/mcp-server"]
}
}
}
# Or: Settings > MCP > Add Server
# Then use mcp_login to authenticate
# Cline: MCP Servers > Configure MCP Servers
{
"mcpServers": {
"jaumemory": {
"type": "stdio",
"timeout": 60,
"command": "npx",
"args": ["-y", "@jaumemory/mcp-server"]
}
}
}
# Then use mcp_login to authenticate
# Windsurf Settings > Cascade > Add MCP Server
# Or edit ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"jaumemory": {
"command": "npx",
"args": ["-y", "@jaumemory/mcp-server"]
}
}
}
# Then use mcp_login to authenticate
# VS Code 1.102+ with Copilot
# Create .vscode/mcp.json
{
"mcp": {
"servers": {
"jaumemory": {
"command": "npx",
"args": ["-y", "@jaumemory/mcp-server"]
}
}
}
}
# Then use mcp_login to authenticate
# ChatGPT Plus/Pro required
# Settings > Connectors > Developer Mode
# In chat: "+" > Developer Mode > Add sources
{
"name": "JauMemory",
"description": "Persistent shared agent memory",
"mcp_server_url": "https://mem.jau.app/mcp/<mcp_id>",
"authentication": "OAuth"
}
# Find your mcp_id in User Dashboard > MCP
# Add to ~/.codex/config.toml
[mcp_servers.jaumemory]
command = "npx"
args = ["-y", "@jaumemory/mcp-server"]
# Then use mcp_login to authenticate
Private beta
JauMemory is invite-only while we harden the platform with early users. Tell us how you use AI, and we'll send an invite as capacity opens up.
Have an invite code? Sign up
Not ready to apply? We'll email you when open signup launches. No newsletter, no spam.
FAQ
MCP (Model Context Protocol) is an open standard that lets AI applications connect to external tools and data. JauMemory is an MCP server, so any MCP-capable client can read and write the same memory: no plugin per app, one protocol for all of them. More →
Claude Desktop, Claude Code, Cursor, Cline, Windsurf, GitHub Copilot, and ChatGPT all work today. Anything else that speaks MCP can connect with the same one-line config. More →
Yes. Your memories are encrypted at rest with per-user AES-256 keys, protected by TLS in transit, and isolated with row-level security in the database. Search runs on our servers with local embeddings, so your content is never sent to a third-party AI API. You can export your data at any time. More →
It's free during the private beta. Paid plans open when billing launches, and beta users will get notice before anything changes. More →
Not currently. JauMemory is a cloud-hosted service at mem.jau.app, and the hosted product is not available for self-hosting today.
Agents are named identities for your AI sessions: each gets its own memory, can share context with others, and learns from recorded errors. Collections are folders for memories, so you can group work by project or topic and search within just one. More →