Skip to content

x-thread-cross-device-memory.md

Latest commit

 

History

History
121 lines (72 loc) · 2.55 KB

x-thread-cross-device-memory.md

File metadata and controls

121 lines (72 loc) · 2.55 KB

X Thread: How to make your Claude's memory work across all your devices

Tweet 1 (Hook)

Your Claude forgets everything when you switch devices.

Your MacBook Claude doesn't know what your iMac Claude learned yesterday.

I fixed this. Here's how to give Claude cross-device memory in 2 minutes:

Tweet 2 (Problem)

The problem:

  • Claude Code stores memory per machine (~/.claude/)
  • Switch laptop? Start from zero
  • Different project on different device? No shared context

Your AI should remember YOU, not just your current machine.

Tweet 3 (Solution intro)

I built Cortex — a local memory engine that syncs through YOUR cloud storage.

  • iCloud, Google Drive, OneDrive, Dropbox
  • AES-256-GCM encrypted (even if your cloud is hacked, memories stay private)
  • 156us writes. 568us search. Pure Rust. 3.8MB.

Tweet 4 (How it works)

How sync works:

Mac A → writes to iCloud/cortex-sync/
Mac B → reads from iCloud/cortex-sync/

No server. No API. No account.
Just your own cloud folder.

Each device writes its own append-only log. No conflicts ever.

Tweet 5 (Setup)

Setup (2 minutes):

# Install
curl -fsSL https://raw.githubusercontent.com/gambletan/cortex/main/install.sh | bash

# Register as Claude MCP server
claude mcp add cortex --scope user -- ~/.local/bin/cortex-mcp-server ~/.cortex/memory.db

Then just tell Claude: "Enable cross-device sync"

Claude calls sync_enable, auto-detects your cloud provider, done.

Tweet 6 (What happens next)

Now when you chat with Claude on Device A:

"I prefer dark mode, use Neovim, and my timezone is Asia/Shanghai"

Switch to Device B, Claude already knows. No setup. No copy-paste. Just works.

Beliefs, facts, preferences, people graph — all synced.

Tweet 7 (Privacy)

Privacy first:

  • Private memories (default) NEVER leave your device
  • Only Shared/Public memories sync
  • End-to-end encrypted with your passphrase
  • Zero telemetry. Zero cloud servers. Verify: grep the source code yourself.

Your memories are yours. Period.

Tweet 8 (Benchmarks)

vs Mem0 cloud:

  • 528x faster
  • Free (Mem0: $99/mo)
  • 100% local
  • 4-tier memory (Mem0: flat)
  • Bayesian beliefs, people graph, contradiction detection

LoCoMo benchmark (ACL 2024): Cortex 73.7% vs Mem0 66.9% vs OpenAI Memory 52.9%

Tweet 9 (CTA)

Cortex is MIT licensed, 100% open source.

Give it a try. Give it a star if it helps.

github.com/gambletan/cortex

Now in English, Chinese, Japanese, and Korean.

Hashtags

#Claude #ClaudeCode #AI #MCP #Rust #OpenSource #AIMemory #Privacy