How to keep a Cursor project plan across chats
Models do not remember between chats, and Cursor rules exist to carry reusable instructions into each one. A plan changes every day, so keep it on a Develosaur board connected to Cursor over MCP, and add one Always Apply rule telling the agent to read the board first. Every new chat then starts from the current plan.
Steps
- Add this to ~/.cursor/mcp.json for all projects, or .cursor/mcp.json in one project:
{ "mcpServers": { "develosaur": { "url": "https://mcp.develosaur.com/v1/user/mcp" } } } - Open Cursor Settings, MCP, and enable develosaur. Cursor opens the sign-in in your browser when the server asks for it.
- Ask the agent: "Read my develosaur board digest and tell me what is next."
- Add a project rule set to Always Apply with this text:
At the start of each session, read the develosaur board digest and continue from it. Claim a task before working on it, and hand off at the end.
Try it without an account
Paste this into your agent. It builds a temporary board you can keep later:
Read https://mcp.develosaur.com/llms.txt and use it to build me a develosaur board for my projectQuestions
Why not keep the plan in a rules file?
Rules suit instructions that rarely change. A plan changes as work gets done, and several chats or agents need to update it without anyone editing a file by hand.
Can Cursor and Claude Code share the plan?
Yes. Both connect to the same board, claim different tasks and see each other's changes live.
Can I try it without an account?
Yes. Paste this into an agent that can fetch URLs: "Read https://mcp.develosaur.com/llms.txt and use it to build me a develosaur board for my project". It builds a trial board you can keep later.
Sources
Related
Other answers
- How to make Claude Code remember the plan between sessions
- How to run multiple Claude Code agents without conflicts
- How to turn a design doc into tasks with Claude
- How to share an AI agent's plan with your team
- MCP task manager for Claude Code: how to set one up
- How to put Codex and Claude Code on the same plan
- How to hand off work between AI agent sessions
- A kanban board your AI coding agents can update
- Persistent memory for AI coding agents vs a project board