🧠 GrayMatter

Agent memory traces — visualized as an interactive 3D brain.

GrayMatter lives in your editor as a raymarched 3D brain. Every memory your AI agent stores gets mapped to an anatomical region — rotate, zoom, and click to explore what it knows.

What it does

GrayMatter gives your AI agents a persistent, inspectable memory store — and lets you see it as a living brain.

• 3D Brain Visualization — WebGL raymarched brain with halftone dot-art rendering, sparkle effects, and smooth rotation/zoom
• Six Memory Regions — Decisions & Planning, Technical Knowledge, Events & History, Visual & Design, Habits & Workflows, Preferences & Style
• Auto-Categorization — Add a memory and Claude sorts it into the right region with a summary, tags, and importance score
• Click to Explore — Click a category or brain region to browse stored memories; the brain rotates to face the selected region
• Brain Health Score — A live badge shows how well-populated your brain is across all six regions (Sparse → Growing → Healthy → Excellent)
• Forget — Hover any memory card and click "Forget" to remove it (with confirmation)
• Import / Export — Back up or share memories as JSON; import from Claude Code, ChatGPT, Cursor, Copilot, and generic markdown formats
• Resizable Panels — Drag the grip between the brain and memory list

Multi-Agent & Multi-Model Workflows

Cross-Agent Memory Sync GrayMatter watches memories.json for external writes. Any agent or tool that writes to the same storage path — Claude Code, Cursor, a custom script — automatically triggers a live reload in the brain view. Set a shared path in settings to give every agent in your stack access to the same memory store.

Memory-Augmented Context Injection Run GrayMatter: Copy Relevant Context (or select text in the editor first) to rank your stored memories against what you're currently working on and copy the top matches as a formatted prompt block to your clipboard. Paste into any LLM chat to give it instant project context.

Contradiction Detection Every new memory is automatically checked against existing ones by a fast model. If a conflict is detected, a warning appears with the explanation and a Resolve action — which sends both memories to a stronger model for a resolution suggestion.

Memory Distillation Run GrayMatter: Distill Memories to send your full memory store to Claude. It identifies redundant or overlapping entries and merges them into tighter, higher-signal memories — keeping your store lean without losing information.

Model Attribution Every memory is tagged with the model that created it. Attribution badges appear on memory cards in the brain view so you can trace memories back to their origin.

Context File Auto-Detection On first open, GrayMatter scans your workspace for known AI context files (CLAUDE.md, .cursorrules, AGENTS.md, MEMORY.md, .github/copilot-instructions.md, and more) as well as any .md file in the project — design docs, general docs, anything markdown. If found, it offers a one-click import to populate your brain immediately — no manual setup needed.

Proactive Memory Surface A status bar indicator (⚡ N memories) lights up whenever you switch to a file that matches memories in your store. Click it to see a ranked list of relevant memories for your current context — without leaving the editor.

Team Brain Export Mark individual memories as shared (⬆ Share button on the memory card) and run GrayMatter: Export Team Brain to snapshot only those memories into a shareable JSON. If nothing is shared yet, you can export everything. Import the snapshot on another machine to share project context across your team.

Sample Memories on First Launch If you skip importing a context file during onboarding, GrayMatter seeds six sample memories covering all brain regions so you can immediately see and explore a populated brain — delete them whenever you're ready to start fresh.

Installation

VS Code

Available on the VS Code Marketplace — search "GrayMatter Memory" in the Extensions sidebar.

Or install a .vsix manually:

code --install-extension graymatter-0.1.2.vsix

Open VSX (Gitpod, VSCodium, etc.)

Available on the Open VSX Registry — search "GrayMatter Memory" in the Extensions sidebar.

Usage

Open the GrayMatter panel from the Activity Bar — or use the command palette:

Click a category tab at the top or a region on the brain to filter memories. Drag to rotate the brain, scroll to zoom. Use Copy Relevant Context before any LLM chat to inject what GrayMatter knows about your current task.

Configuration

Building from Source

git clone https://github.com/lucilehan/graymatter.git
cd graymatter
npm install
npm run compile

To package:

npx @vscode/vsce package --no-dependencies

Contributing

Contributions are welcome! Please read CONTRIBUTING.md before opening a pull request.

• Bug reports / feature requests — open an Issue
• Code changes — fork → branch → PR against main

Changelog

0.1.3

• Proactive Memory Surface — status bar shows relevant memories for the active file; click to browse
• Brain Health Score — live badge (Sparse → Excellent) tracks coverage and depth across all six regions
• Team Brain Export — mark memories as shared and export a team snapshot with Export Team Brain
• Broader .md scanning — auto-detection now picks up ALL markdown files in the project (design docs, general docs, etc.) not just known filenames
• Sample memories on onboarding — skip import and get 6 demo memories covering all brain regions immediately
• Enhanced onboarding modal UI with step indicator and import drop zone

0.1.2

• Added VS Code Marketplace and Open VSX badges and install links to README
• Made Open VSX publish non-blocking in CI

0.1.1

• Initial public release on VS Code Marketplace
• Renamed display name to GrayMatter Memory to avoid marketplace conflict
• Dynamic GitHub Release titles using latest commit message

0.1.0

• First release: 3D WebGL brain, six memory regions, auto-categorization, contradiction detection, memory distillation, context injection, multi-agent sync, import/export

Made with 🍸 by lucilehan