A2CR MCP server

A2CR is a small helper that lets your AI save a short summary of what you were working on, then hand that summary to a new chat window later. It is handy when a long task fills up your chat and you want to start fresh without losing the thread. It works with AI clients like Codex, Claude Code, and Cursor.

What this MCP server does

You ask your AI to save where things stand, and it writes a short checkpoint called a WorkBaton into a small database on your computer. Later, in a brand new chat window, you ask it to resume, and it reads that checkpoint back so the new window knows the goal, the current state, and the next step. If you have longer notes, it can keep those separately as WorkStash notes and point the checkpoint at them. Everything stays on your machine, and nothing is sent to an A2CR server.

Click to zoom

What you can do with it

• Save a short checkpoint of your current work
• Resume that work in a fresh AI window
• List the checkpoints you have saved
• Load one specific checkpoint by name or number
• Delete a checkpoint you no longer need
• Store a temporary supporting note and retrieve it later
• Ask the AI whether saving a checkpoint makes sense right now

Try asking your AI

• “Save a checkpoint of where we are so I can start a new chat later.”
• “Resume my last checkpoint and tell me what the next step was.”
• “List all my saved checkpoints.”
• “Store a note about why we changed the login logic, then reference it in the checkpoint.”

What it gives back to you

You get short, readable answers in the chat. When you save, it confirms the checkpoint was stored and often shows the goal and next step it wrote down. When you resume, it reads the checkpoint back to you as a summary of the goal, current state, and next action. Listing gives you a simple list of your saved checkpoints.

Before you start

Who it's for

People who run long AI coding or writing tasks and want to move the work into a fresh chat without re-explaining everything.