One memory for every AI
One memory for every AI you use.
Tell Entity Memory something once in Claude, and it is already known in ChatGPT, Cursor, and every other tool you open. Save a correction in one. It follows you into the next, in every session, on every machine. No more starting from zero. No more re-explaining yourself.
Always uses pnpm, never npm.
saved in Cursorrecalled in Claude
See it in one place
Everything your tools remember, in one home.
Always uses pnpm, never npm. No AI co-author trailer in commits.
Prefers concise, direct answers. Leads with the recommendation, no theory dumps.
Based in Fremont, California. Co-founded a robotics company, exited to Miko.
One account. Every tool.
Every tool reads and writes the same memory. Connect with one link, or install locally in one command.
and any tool that speaks the open MCP standard.
Without a shared memory
You keep introducing yourself to tools that should already know you.
Claude
use pnpm, not npm
starts from zero every session.
Cursor
writes concise, no dashes
never heard your rules.
ChatGPT
married to Aishani
asks you to explain it again
Fix something once and the others already know. No re-explaining, no copy-paste between tabs.
Up and running in two minutes
How it works
- 01
Connect a tool.
One sign-in link for the hosted cloud, or one init command for a local server.
It auto-detects Claude Code, Cursor, Windsurf, and Cline, and writes their config for you.
npx entity-memory init - 02
It remembers as you work.
Corrections, preferences, and project facts are captured in the flow.
No extra step, no tagging, nothing to maintain.
- 03
It follows you to the next tool.
Open the next assistant and your context is already there.
The fix you made in one is known in all.
From the first message
A brand-new chat already knows who you are.
Your identity, your preferences, the way you work, the people who matter, and the rules you never want repeated. A fresh assistant opens already holding them, so you skip the introductions and get straight to the work.
Corrections and preferences rise to the top. One-off notes settle down. Every result arrives with a plain-language reason it is relevant, not a keyword dump.
surfaces the right memory, ranked for you.
Profile snapshot
- Identity
- Founder, building in AI and consumer hardware
- Preferences
- Concise, direct, no dashes
- Work style
- Ship ready-to-use outputs, think in systems
- Relationships
- Spouse, co-founder, key teammates
- Durable rules
- Always run tests after changes
Prefers concise, direct copy with no dashes.
Always uses pnpm, never npm.
Text, images, voice, PDFs
Find it even when you only remember the gist.
One plain question surfaces the screenshot, the voice note, and the spec, all in one search space. No tagging, no folders. Hybrid search blends meaning and exact words, so you are covered whether you recall the phrase or just the idea.
Drop in a folder of documents and ask any of your assistants about it.
Typed, ranked, de-noised
Not every note deserves the same shelf.
Ranked recall
The few facts that answer the question. Recall re-ranks so corrections and preferences rise and raw chunks sink, each result carrying a reason it is relevant.
Self-maintaining
Clean on its own. Repeats are reinforced, stale facts are replaced, duplicates never pile up. You never tidy it.
Rules stay put
The important rules stay sharp. Corrections never fade and rank first. One-off notes gently settle over time.
Always uses pnpm, never npm.
Tried a coffee shop in Lisbon once.
Yours, on your terms
Built for how you work, and who you work with.
Per-tool workspaces
Run Claude on personal and Codex on a team workspace at once. Switch or revoke any tool's access in one click from Settings.
Team sharing that compounds.
One teammate fixes a mistake once, and every teammate's assistant stops repeating it. Isolation is enforced in the database itself, not just in app code, so outsiders see nothing.
You pick which tool sees which workspace.
Revoke any tool in one click.
Open source and self-hostable. Yours to keep.
Works offline on the local setup.
Think of it as iCloud for your tools. You set it up once, and your memory syncs everywhere you work.
Beyond the apps
Your own bots share the same memory.
A stable REST API and an experimental Python SDK expose the same data, search, and workspaces the apps use, so a custom assistant or a WhatsApp bot reads one source of truth.
Sutra Bot, a WhatsApp tutor over a roughly one-million-word scripture corpus, runs on Entity Memory.
from entity_memory import Client
em = Client(api_key="em_...")
em.remember("Always uses pnpm, never npm.")Sutra Bot · proof
Retrieves a cross-language scripture passage and renders it back in the reader's language, grounded in one shared memory.
Open, self-hostable, yours.
Open source
Apache 2.0
yours to fork
Self-host
1 Supabase + 1 Vercel
your data, your infra
Local install
1 command
auto-detects 4 tools
Connect
1 sign-in link
no keys to juggle
From the founders of Square Off.
Questions, answered plainly.
No. Connect with one sign-in link, or install locally with one command. Entity Memory writes the config for Claude Code, Cursor, Windsurf, and Cline for you.
Entity MemoryOne memory for every AI you use.
Connect your first tool in two minutes.