Status · Open source

Backend · File-based · QMD optional

Ships with · OpenClaw 2026.2.6+

Pattern · Dory-Proof

Justin Hart / Systems & IP / Dory Memory System

Open source · System & IP

Teach AI to remember.

A file-based memory system for AI agents that forget between sessions. Named after the forgetful fish — because your AI is brilliant, but it's also a goldfish.

Install the skill Read the article →

workspace · file tree

● live

# Your agent reads this every session
workspace/
├── SOUL.md          # Who the agent is
├── USER.md          # Who you are
├── MEMORY.md        # Long-term memory
│
├── state/           # Check every session
│   ├── ACTIVE.md    # Current task (exact words)
│   ├── HOLD.md      # Blocked items
│   └── DECISIONS.md # Recent choices
│
└── memory/
    └── YYYY-MM-DD.md

01 · The pattern

The Dory-Proof pattern.

Capture before interpret · Always

— The rule

When the user gives a task, immediately write their exact words to a file. Then interpret. Then act.

Not a paraphrase. Not a summary. The exact words. Because interpretation drifts, but quotes don't.

This is the difference between an agent that ships what was asked for, and an agent that ships what it thought was asked for. The cost of writing one extra file is microseconds. The cost of drift is a wasted session.

02 · The principles

Four rules that hold
the system together.

All four · or none of them

Text > brain.

Files ARE memory. If it's not written down, it doesn't exist.

ii.

Structure prevents chaos.

HOLD overrides ACTIVE overrides DECISIONS. Clear hierarchy, every time.

iii.

Exact words matter.

Capture user's exact words before interpreting. Prevents drift.

iv.

Boot sequence.

Every session reads state files first. Continuity guaranteed.

03 · Level up

QMD: semantic search
built in.

Ships with OpenClaw 2026.2.6+

● Built-in to OpenClaw

Local semantic search
over your workspace.

QMD auto-indexes your workspace every 5 minutes. Searches by meaning, not just keywords. Session transcripts included.

The Dory-Proof pattern remains the foundation. QMD adds intelligent retrieval on top — so your agent doesn't just remember, it can find what it remembered.

openclaw.json

// Enable QMD in your config
{
  "memory": {
    "backend": "qmd",
    "qmd": {
      "update": { "interval": "5m" },
      "sessions": { "enabled": true }
    }
  }
}

— The thesis

"The Dory Problem isn't about AI having bad memory. It's about us expecting AI to work like humans when it doesn't."

From the original article — Hello, AI

04 · Get started

Install the skill.
Start remembering.

Drop the skill into OpenClaw, copy the templates, point your agent at the workspace. The boot sequence does the rest.

Install the skill Read the full article →