Project Memory Management Tool 'mex' for AI Agents Released

A tool called mex has been released to address the issue where coding agents make errors by relying on outdated or incorrect memory (context) when a project's structure or dependencies change.

mex creates a Markdown wiki in the .mex/ directory containing architecture, conventions, and decisions, providing agents with persistent project memory. Key features include:

• mex check: Uses a deterministic method without LLMs to verify discrepancies by comparing memory content against the actual file system, scripts, and dependency manifests, and calculates a health score.
• mex sync: Implements a 'check → targeted repair → verify' loop that targets only files requiring fixes based on verified errors and requests corrections from the agent.
• Code Graph Layer: Links Markdown knowledge to actual code symbols to ensure knowledge persists even when implementations change or move.