PML — Project Memory Layer
>
Continuous project authority for humans + AI
PML is a lightweight system that
remembers your project so you don’t have to — and makes sure AI tools don’t forget either.
It watches your codebase, tracks decisions, enforces rules, and continuously feeds
correct context to tools like
Cursor and
GitHub Copilot.
No dashboards.
No tickets.
No ClickUp clones.
Just
memory, authority, and flow.
---
Why PML Exists
Modern development is broken in a subtle way:
- AI can scaffold fast, but
forgets context
- Rules live in your head, not the repo
- Decisions disappear after one prompt
- Folder structures slowly rot
- AI tools confidently violate constraints
You end up:
- Re-explaining the project
- Fighting hallucinations
- Writing longer prompts
- Losing trust in automation
PML fixes this by becoming the project’s long-term memory.
---
What PML Does
$3
Stores:
- Project identity
- Folder structure
- Rules & constraints
- Architectural decisions
All in a single file:
pml.json
---
$3
PML runs quietly in the background and:
- Watches file changes
- Detects structure drift
- Regenerates AI instructions automatically
It doesn’t interrupt —
unless something matters.
---
$3
PML doesn’t just observe. It enforces.
- New folders? → flagged
- Forbidden libraries? → violations
- AI drift? → corrected at the source
This turns your project from
suggestion-based to
rule-based.
---
$3
PML generates and maintains:
-
.cursor/rules.md
-
.github/copilot-instructions.md
So your AI tools always receive:
- Correct structure
- Current rules
- Real decisions
No more “please remember…” prompts.
---
Installation
npm install -g pml
or clone and run locally during development.
Quick Start
Initialize memory
pml init
Creates pml.json in your project.
Sync project structure
pml sync
Scans folders and updates memory.
Remember a rule
pml remember "Never use shadcn in this project"
Rules become enforceable constraints.
Explain context (for humans or AI)
pml explain auth
Outputs a clean, focused context snapshot.
Copy AI-ready context
pml context auth
Copies structured project context directly to clipboard.
Run the daemon
pml daemon
PML now watches your project continuously.
Example output:
● MODIFY src/core/context.ts
✓ Project structure synced
✓ Cursor rules updated
✓ Copilot instructions updated
Violations (The Important Part)
PML flags problems as they happen:
✗ VIOLATION
Rule: Undeclared folder
File: src/random/experiment
Future versions will support:
Rule approvals
Auto-fixes
Commit blocking
AI response validation
What PML Is Not
❌ A task manager
❌ A PM tool
❌ A Notion replacement
❌ A UI-heavy SaaS
PML lives inside your workflow, not on top of it.
Philosophy
Projects need memory.
AI needs authority.
Humans need flow.
PML sits between all three.
Status
Core CLI: ✅
Daemon: ✅
Cursor & Copilot sync: ✅
Structure violations: 🚧
Rule enforcement: 🚧
AI drift detection: 🚧
This project is early, opinionated, and moving fast.
License
MIT.
Use it. Break it. Improve it.
Final Note
If you’ve ever thought:
“Why does my AI forget everything I just explained?”
PML is for you.