Skip to content

MEMORY FOR EVERY AI

Your notes are
your AI's memory.

Frontier-grade memory for your own notes, reachable by every AI you use, even with your laptop closed.

en·gram: a unit of cognitive information imprinted in a physical substance.

HOW IT WORKS

You bring the notes.
We handle the hard part.

No setup, no upkeep, no lost weekend getting it working. Three steps, then your AI just knows.

  1. Step 01

    Bring your notes.

    Plain markdown in a folder you own. No proprietary format, no lock-in. Already use Obsidian? Point Engram at your existing vault. Starting fresh? Open the web app and write.

  2. Step 02

    Connect any AI.

    Claude, Cursor, ChatGPT, or whatever you use. Connect once, and your AI can read your notes, write new ones, and update what's already there.

  3. Step 03

    Ask anything.

    Your AI pulls the right notes itself. No more pasting context. No more "what did we decide last quarter?" It already knows, because it can read what you wrote.

Switch AIs.
Keep your memory.

The best AI changes every month. With built-in assistant memory, every switch means starting over: re-explaining everything to a model that has never met you. Engram keeps your AI's memory in your hands instead, so you can switch whenever you like and the new one already knows you.

  • No vendor lock-in.

    Your memory lives in plain markdown over an open standard, not sealed inside one assistant. Leave any time; lose nothing.

  • Every AI, one memory.

    Claude today, whatever's better tomorrow. They all read and write the same vault. Your context compounds instead of resetting.

  • Never introduce yourself twice.

    The notes you already wrote are the context. New model, new app, new device: it picks up where the last one left off.

INSTANT CONTEXT

Ask.
It already knows.

Skip the re-explanation. Your AI pulls from notes you already wrote and answers like it remembers. Because, through your notes, it does.

You: "What were my main objections to the Anderson paper?"

Claude: Searched your vault, found /reading/anderson-2025.md and 2 linked notes…

"You flagged the sample size as too small to generalize, and that the control group wasn't blinded. You wanted to chase down their appendix data."

It searches by meaning, not just words. Ask for "our refund policy" and Engram surfaces the note that never says the word "policy." Then your AI answers from it, like it remembered.

Works with

  • CLAUDE
  • CLAUDE CODE
  • CURSOR
  • CHATGPT (WEB)
  • CLINE
  • CONTINUE
  • COPILOT
  • WINDSURF
  • ANY MCP CLIENT
See all integration guides
ONE VAULT, EVERYWHERE

One vault.
Edited from anywhere.

A vault is a folder of markdown notes you own. Engram syncs it between every device you write on (desktop, mobile, tablet, browser), and while it syncs, it indexes every word so any AI you connect can recall it.

  • In Obsidian, on every device.

    Desktop, mobile, tablet: your vault stays in sync without iCloud hacks or a separate Obsidian Sync subscription. Same notes everywhere you open them.

  • In the browser.

    app.engram.page: open and edit on any machine, nothing to install.

  • By your AI.

    Claude, Cursor, ChatGPT: every AI you connect reads and writes the same vault you do.

  • Never lose an edit.

    Offline edits queue and replay on reconnect. Nothing is ever silently overwritten.

OBSIDIAN WEB # VAULT CLAUDE CURSOR CHATGPT
  • OBSIDIAN
  • WEB APP
  • MOBILE
  • ALWAYS IN SYNC
WHO IT'S FOR

Four ways in.
One memory underneath.

  • Already deep in Obsidian.

    Years deep in a vault you love. Engram syncs it across every device without the iCloud gymnastics, and makes it readable and writable by every AI you use. Drop-in, no migration.

  • Glued to Claude and Cursor.

    Claude, Cursor, ChatGPT all day, re-pasting the same context every session. Give your AI a memory it can read and write, and keep it when you switch models. Stop re-explaining yourself.

  • Capture everything, reuse nothing.

    A second brain that's write-only. Open the web app, keep writing how you already write, and let your AI actually find and use it. "I know I wrote that down somewhere" becomes "it already knows."

  • Won't touch the cloud.

    You won't put your knowledge on someone else's server. Self-host the source-available backend with one Docker command: full features, full API, your hardware, your rules. Your notes never leave it.

OWNERSHIP

Yours to own.
Yours to leave.

We're straight with you about what this is: notes you own, encrypted at rest, never sold. Not a zero-knowledge vault. Want the strongest privacy posture? Run it yourself. More on self-hosting just below.

Encrypted, with a key that's yours.

Your notes are encrypted on our servers and in transit, under a key tied to you alone. Even your folder names and tags are scrambled, never sitting around in plain text.

Never sold. Never trained on.

Your notes don't feed anyone's model, including ours. No analytics resale, no third-party data brokers, no telemetry by default.

Plain markdown. Walk away anytime.

No proprietary format. Cancel whenever and leave with every note in a folder you can open in any text editor. Owning the exit is the whole point.

Strongest

Want it so even we can't read it? Self-host.

On the hosted service, we process your notes to index them, so we're honest that it isn't zero-knowledge. Run Engram on your own hardware and your notes never leave it. Same features, total control.

  • ENCRYPTED
  • NEVER SOLD
  • NO AI TRAINING
  • NO TRACKING
  • WALK AWAY ANYTIME
SELF-HOSTABLE

Or run the whole thing
on your own hardware.

Server, sync engine, web app, embeddings: every byte on hardware you control. The strongest privacy posture we offer, free, out of the box. (Unless losing the weekend to a good setup is the fun part. Same. It's all yours.)

One Docker command.

docker compose up brings up the whole stack. No Kubernetes degree required. Runs on a homelab box, a NAS, or a cheap VPS. Just remember to back up your encryption key: lose it and your notes are gone for good.

Source-available. Read every line.

Server, plugin, and web app are public on GitHub under the PolyForm Small Business License, with a commercial license for larger organizations. Personal use and small-business self-hosting are unrestricted. Audit the crypto, fork the sync engine, send a PR.

Local AI, if you want it.

Pair Engram with Ollama or any OpenAI-compatible local model and your vault never leaves the LAN. Pay nobody for embeddings. The only network call your notes make is between two of your own machines.

No phone-home, no kill switch.

Cut the WAN cable and the stack keeps running. No activation handshake, no usage beacons, nothing we can flip off remotely. Air-gap it if you want.

  • DOCKER COMPOSE
  • SOURCE-AVAILABLE · POLYFORM SBL
  • LOCAL EMBEDDINGS
  • OLLAMA COMPATIBLE
  • AIR-GAP READY
View the source on GitHub
PRICING

Free to start.
Pay when you scale.

Monthly or annual. Annual saves ~17%. Self-hosting is free, forever.

  • Free

    $0

    • 1 vault · 1 device
    • 10k notes · 1 GB
    • Search by meaning + keyword
    • Connect any AI
    • 5 AI conversations/day
  • Starter

    $7/mo

    or $70/year

    • 5 vaults · unlimited devices
    • 50k notes · 3 GB
    • Real-time sync
    • Connect any AI · full API
    • 500 AI queries/day
  • RECOMMENDED

    Pro

    $14/mo

    or $140/year

    • 15 vaults · unlimited notes · 16 GB
    • 10,000 AI queries/day
    • Better-ranked results · search all vaults at once
    • Highest usage limits
    • Priority support
See full comparison
FAQ

Common questions.
Straight answers.

  • Do I need Obsidian?

    No. The web app at app.engram.page is a full editor. Sign in and write. Obsidian is a great option if you want a desktop-native experience: install the plugin and your existing vault syncs, no migration.

  • Which AI tools does it work with?

    Any client that speaks MCP, the open Model Context Protocol: Claude, Claude Code, Cursor, ChatGPT (web), Cline, Continue, GitHub Copilot, Windsurf, and more, with step-by-step guides for each. If your client speaks MCP, it works.

  • Is it end-to-end encrypted?

    Honestly: no, not on the hosted service. Your notes are encrypted on our servers and in transit, and we never sell them or train AI on them. But we do process them to make them searchable, so we're not zero-knowledge. If you want notes that never leave your hardware, self-host. Then they're entirely yours.

  • Can I import my existing notes?

    Yes. Drop a folder of markdown into the web app, or point the Obsidian plugin at an existing vault. Engram indexes in the background while you keep working.

  • What if my AI writes the wrong note?

    Every AI write is logged with a timestamp and source, and nothing is ever silently overwritten. The web app shows per-note history with one-click revert, so you're never stuck with a mistake.

  • Does it work on mobile?

    Yes. The Obsidian plugin runs on iOS and Android, and the web app works in any modern mobile browser. Your vault stays in sync across every device.

  • What does it cost?

    There's a free tier to start. Paid plans are Starter at $7/mo ($70/yr) and Pro at $14/mo ($140/yr). Self-hosting is free.

EARLY ACCESS

Stop repeating.
Start remembering.

Join the waitlist and we'll email you a one-click checkout link the moment Engram opens.

By joining, you agree to our Privacy Notice. We collect your email to send launch updates. We don't sell or share personal information.

One confirmation email, then nothing until launch. Unsubscribe any time.