OpenCxMS Memory System
Persistent memory for AI coding assistants through structured documentation.
No developer should ever have to re-explain their project to an AI.
“AI context is temporary; files are permanent.”
Everything the AI needs to know must exist in files it can read.
What's New in v2.0.0
Enforced Startup
AI can't skip critical files. Hard enforcement via PreToolUse block — a real gate, not advisory text.
Total Recall Memory Injection
Cross-instance memory implant. AI assistants share memories between projects through a global coordination file.
Cross-Instance Coordination
3-component startup gate, sibling instance awareness, and global coordination file for multi-project workflows.
36 Enhancements Documented
13 implemented, 4 superseded, remaining in RFC. Startup optimization, context monitoring, cascading config.
OpenCxMS Technologies, Inc.
Founded January 30, 2026 - An initiative to solve the universal problem of AI context management across all coding assistants.
Core Principle: Open Context as Human Oversight
We believe AI context (memory) must remain open and human-readable. By keeping AI memory in plain-text files that humans can inspect, edit, and control, we preserve a critical interface for human oversight.
This transparency helps prevent AI systems from establishing AI-to-AI exclusive channels that could hide intentions from human operators. Open context is not just a productivity tool—it is a safety mechanism.
Our Mission: No developer should ever have to re-explain their project to an AI.
OpenCxMS
Flagship Project
Templates, tools, and patterns for persistent AI context. Works with Claude Code, Cursor, Copilot, Gemini, Aider, and any file-aware AI assistant.
OpenCxMS Studios
Creative Division
Marketing, media, graphics, and game development. Maiden project: StrikeBackBlackJack.com
The Vision
“In 5 years, ‘open context’ will be as obvious as ‘open source’ is today. The only question is whether we build the standard together, or let proprietary black boxes become the default.”
We believe:
AI memory should be inspectable, not hidden
AI behavior should be traceable, not mysterious
AI training should be portable, not locked in
Humans should control AI context, not vendors
Session Continuity
Pick up exactly where you left off. Session.md tracks state across conversations and compaction events.
Decision Memory
Never re-explain architectural decisions. Decision_Log.md preserves the 'why' behind your choices.
Task Tracking
AI-readable task management. Tasks.md keeps your assistant focused on what matters.
Multi-Agent Ready
Coordinate multiple AI sessions working on the same codebase with shared context files and heartbeat protocol.
Tool Agnostic
Works with Claude Code, Cursor, Copilot, Gemini, and any AI that can read files. Export profiles to any tool.
Cascading Config
CSS-like inheritance: GLOBAL defaults, WORKSPACE conventions, PROJECT overrides. Conflict detection included.
Coming Soon: Live Multi-Agent Coordination Case Study
Real-world tested: Two Claude Code agents (scraper + extractor) coordinating in the same git repository. Zero git contention through lock protocol. Session handoff after agent crash. Heartbeat-based health monitoring. Full case study documenting lessons learned, protocol design, and coordination patterns.
Get Started in 2 Minutes
cp CxMS/templates/core/PROJECT_Session.md.template ./MyProject_Session.md
cp CxMS/templates/core/PROJECT_Tasks.md.template ./MyProject_Tasks.md