Latent memory — agents plant the cards. You tend the garden.

You studied merge sort last week.

A week on, it’s already fraying.

Gone. That’s the forgetting curve.

Recall it right before it slips.

Each recall holds longer than the last.

Weeks apart now — and it stays.

scroll

You never write a card. They plant themselves while you work.

Days later, mid-conversation — it asks.

You answer. It schedules the next ask.

Why is Postgres skipping my index on email?

Your query filters on lower(email) — a plain index can’t serve that. You want an expression index.

create_card · “when does Postgres skip a b-tree index?”

six days later, different conversation

get_due_cards · 1 due

Quick one while we’re here — why would Postgres skip that index on email?

The query wraps the column in a function — needs an expression index.

grade_card 5 · next due in 12 days

Yours now. I’ll ask again right before it slips.

Months of conversations. Every card your AI planted.

You only review a handful.

Each strong card lights up its neighbors.

The whole garden stays green.

Your cards don’t live in a tool. Today you chat in Claude.

Next year, who knows — swap freely.

Anything speaking MCP plugs in. The garden stays.

latentmemory

Agents plant the cards.

You tend the garden.

$ claude mcp add --transport http latentmemory https://api.latentmemory.app/mcp

free · sign in with GitHub

Then just keep working. Cards plant themselves — and when one’s due, it finds you.

On claude.ai instead? Settings → Connectors → same URL