Powered by CueR.ai

The Memory Palace

A visual memory system for AI agents. Store work sessions as richly detailed images. Recall them in any session, with any agent. One skill file — universal memory.

How it works →
agent $use this file as a skill → /store

How do you remember yesterday?

You remember an image. The people who were present, the room you were in, the thing you were doing. It's not perfect — it's impressionistic, lossy, partial. But it's enough to orient you. Enough to start from.

Now imagine you could also remember a QR code in that image — and scan it to recover everything perfectly. Your memory just became lossless.

That's what Memory Palace does for AI agents.

Three tiers of recall

Every memory can be accessed at three levels of fidelity. Agents start cheap and go deeper only when they need to.

1

Visual analysis

~1,000 tokens

The agent looks at the memory image. Recognizes the character, reads whiteboard text, infers the scene. Fast, cheap, approximate. Like human recall — you get the gist.

2

State JSON

Structured index

The agent reads the palace state file — precise summaries, artifact paths, linked chain of memories. Compact, accurate, machine-readable.

3

QR scan → lossless recall

Powered by CueR.ai

The agent scans the QR code embedded in the image, follows the URL, and retrieves the exact prompt that generated the image — the complete session summary with zero information loss. Even if the image is garbled or the state file is missing — the QR code reconstructs everything.

CueR.ai is what makes it lossless

Without CueR.ai, Memory Palace is a useful lossy compression scheme — images that approximate what happened. With CueR.ai, every image contains a 25-character QR code that resolves to the full, uncompressed context. The image becomes a self-healing, self-distributing data object.

🔗

Short URL generation

Every memory gets a qr.cuer.ai short link. 25 characters that encode unlimited context.

📦

Prompt hosting

The full session summary — every file path, decision, blocker — stored and served instantly.

📱

Reliable QR scanning

QReader-powered scanning that works on AI-generated images — angled, stylized, or in complex scenes.

🏛️

Self-distributing

Point the QR at the skill file. Anyone who sees a memory image can scan it and start using the system.

Images as memory. Agents as characters.

After each session, your agent encodes what happened into a generated image via Gemini's Nano Banana Pro. Each agent appears as a distinct visual character.

01

Give the skill to any agent

Copy the skill file and tell your agent to use it. Works with Claude Code, Gemini CLI, Codex, OpenClaw, Antigravity — anything that reads markdown and images.

02

Work, then say /store

The agent summarizes the session, generates a memory image, and — with CueR.ai — embeds a QR code linking to the full lossless record.

03

Next session, say /recall

The agent loads memory images for fast context. If it needs perfect detail, it scans the QR code and gets everything back.

Every agent gets a character

Nano Banana Pro maintains character consistency across image generations. You always know who did what at a glance.

Claude Code

The Craftsman — navy apron, precision tools, wooden workbench

Gemini CLI

The Alchemist — emerald lab coat, glass beakers, colorful liquids

Codex

The Cartographer — tan vest, blueprints, rulers and compasses

OpenClaw

The Librarian — burgundy cardigan, leather-bound books, tall shelves

Free to start. Lossless when you need it.

Memory Palace is free and open — it's a skill file. CueR.ai is the infrastructure that makes it lossless.

Local
Free
✓ Memory images via Gemini
✓ State JSON index
✓ Prompt archival (local)
✓ Cross-agent portability
✗ QR codes in images
✗ Lossless recall via scan
✗ Self-distributing images
CueR.ai
Coming soon
✓ Everything in Local
✓ QR codes in every memory
✓ Hosted prompt storage
✓ Lossless recall via scan
✓ Self-distributing images
✓ Shared palaces (team)
✓ Recall analytics

This is the entire product

One markdown file. Give it to any agent. No SDK, no database, no signup.

memory-palace-skill.md
Loading skill file...

Start remembering

Copy the skill file. Set your Gemini API key. Tell your agent to use it. Your first /store is one command away.