MCP Server
Claude, Cursor, Windsurf, and OpenClaw
PRIVATE AI MEMORY ARCHIVE
Open the version you own.
stored locally · powered by Venice · owned by you
PORTABLE BY DEFAULT
One local context layer sits beneath every connected tool.
portable note / context travels without platform lock-in.Claude, Cursor, Windsurf, and OpenClaw
ChatGPT, Gemini, and Perplexity
Local applications
Manual memory management
shared memory rail
One owned context surface beneath every tool.
ARCHITECTURE
Deja Vu is designed so memory extraction, storage, and retrieval do not require handing your complete context to a platform-owned memory system.
local ownership note / extraction, storage, and retrieval stay user-owned.
Forked from Mem0 - entity recognition, deduplication, conflict resolution, and multi-signal retrieval.
EXTRACTPRIVACY MODEL
Deja Vu stores memories locally and uses Venice for zero-retention inference, making privacy a structural property rather than a policy promise.
Provider-owned
Cloud pipeline
Venice - zero retention
Provider cloud
Cloud default
Local SQLite on device
ChatGPT only
Multi-model
All major AI tools
No
Limited
Yes - raw memory file
Policy-dependent
Varies
No - Venice forgets
No
Partial
Yes
No
No
Yes - local-first
WHAT MAKES DEJA VU DIFFERENT
The longer you use Deja Vu, the more useful your AI context becomes - while remaining portable, inspectable, and yours.
compounding note / context grows more useful without giving up ownership.
Works across ChatGPT, Claude, Gemini, Cursor, Perplexity, and open-source models.
Network Utility
$DEJAVU aligns shared infrastructure with local ownership.
OPEN SOURCE
Development happens in public through the GitHub repo, with updates shared on X.