The next AI starts blind
It does not know the current task, the branch, or the last decision unless you teach it again.
Mindswap keeps project context inside the repo so you can switch between AI tools without rebuilding the whole conversation, re-typing decisions, or asking the model to rediscover the work.
The pain is usually the same: a fresh session starts blind, decisions get repeated, and the next tool has to guess instead of continuing.
It does not know the current task, the branch, or the last decision unless you teach it again.
Moving from one editor or client to another often drops the working memory that was already built.
Decisions, blockers, and handoffs live in different places, so the next session has to search and reassemble them.
One repo-level system that saves context, generates handoff files, and gives AI clients a clean way to read the same work.
Set up a repo, save the current state, check health, resume work, and keep the handoff files fresh.
Open CLI guideLet Claude, Cursor, Codex, and other clients ask the repo for context, memory, and history directly.
Open MCP guideUse the package from the official npm registry, then run the CLI commands locally in your repo.
Open npm packageKeep the sequence simple: install, initialize, then choose CLI or MCP depending on how you work.
Clone the repo, install mindswap from npm, and run `npx mindswap init` inside your project.
Step 2Save the current state, check health, and resume later without rebuilding the work from scratch.
Step 3Let an AI client read the repo directly, then use HTTP only when the client is hosted or browser-based.
Open the npm listing, copy the install command, then choose the CLI guide or MCP guide depending on how you want to use mindswap.