Reusable lessons become active guidance
User success criteria, failure-avoidance rules, and proven approaches are retrieved before meaningful work starts.
TOOL CATALOG
VibeBox
Local-first memory middleware for AI coding agents.
VibeBox keeps durable preferences, project rules, validation habits, failure-avoidance lessons, and successful approaches close to the agent workflow without turning the repository into a memory store.
Why VibeBox Helps
AI coding work becomes stronger when the next session can start from remembered context, not from repeated explanations.
Context stays on your machine
The global VibeBox store remains local by default, with an Obsidian-compatible wiki for review instead of cloud sync.
The agent keeps semantic responsibility
The AI agent interprets the user's intent and submits structured candidates; Core validates, dedupes, links, indexes, and renders them.
What Becomes Easier
VibeBox is most useful when a project has recurring preferences, review criteria, or mistakes that should not be rediscovered every time.
Memory in the workflow
The agent retrieves context before work and captures meaningful lessons after the task is done.
Structured memory model
Decisions, preferences, failed attempts, and successful approaches are connected as reusable context.
Readable local wiki
The Obsidian-compatible wiki gives users a reviewable display layer while the local JSON store remains the source of truth.