OP here. We've been building agentic tools and realized we were all solving the same problem: passing context to the LLM. We standardized a folder structure (.ai/) and a "boot protocol" that lets any agent (Cursor, Claude, etc.) read project state from markdown files in the repo, effectively giving the repo its own memory.
We also built a CLI (corepackai) to install these "Context Packs" like npm packages, and a Marketplace to share them (e.g. installing standard Next.js patterns or AWS architecture contexts).
We just released the full spec and would love feedback on the architecture.
OP here. We've been building agentic tools and realized we were all solving the same problem: passing context to the LLM. We standardized a folder structure (.ai/) and a "boot protocol" that lets any agent (Cursor, Claude, etc.) read project state from markdown files in the repo, effectively giving the repo its own memory.
We also built a CLI (corepackai) to install these "Context Packs" like npm packages, and a Marketplace to share them (e.g. installing standard Next.js patterns or AWS architecture contexts).
We just released the full spec and would love feedback on the architecture.