Why Capsules?
Sharing context with AI is usually messy. Pasting raw documentation leads to hallucinated code because LLMs can’t distinguish between a strict rule and a loose suggestion. Capsules solve this by packaging:- Decisions: The strictly enforced architectural constraints.
- Memories: The advisory, explanatory context answering “why something was decided”.
Using Capsules
MCP Server
Load Capsules directly into Claude Desktop via the Model Context Protocol.
IDE Integration
Export to
.contextpack.json format to share with Cursor AI via the IDE addon.