beehiivCursor’s agent shipped three things while you were on a call. The only record is a chat panel you have to scroll back through, in one editor window, on one computer.
Add Context Board to Cursor and the agent can tell your board instead.
What Cursor’s agent puts on the board
Updates: When it finishes a feature, it posts a line to the board’s activity feed. What changed, what it found, what needs a decision.
Its status: While it works, it says what it is on in Agent activity. When it is stuck on a question, that question shows up in Needs you, next to anything else waiting for you.
Your to-do list: It adds the tasks it uncovers and ticks off the ones it finished.
Numbers you care about: It can connect public data the board keeps live by itself, like your repo’s GitHub stars or your package’s weekly npm downloads.
It reads the board, too. Ask what to work on next and it can check your to-do list and the issues on your board before it answers.
What the board sees
Only what the agent sends through the board’s own tools. An update, a status line, a task, a number. Your code stays in Cursor.
Every change shows with Cursor’s name on it, and you can undo it.
Setting it up
In the board’s Connect window, open Your AI, then Cursor, and click Add to Cursor. Cursor opens and asks you to install the board. When it shows Needs login beside it, click that, then Allow on the board’s page. There is no key to copy.
Rather do it by hand? Add the board’s address to ~/.cursor/mcp.json under mcpServers, and Cursor signs in the same way.
If you also run Claude Code or Codex, link your computer with the runner and their sessions appear on the same board without being asked. Cursor’s agent appears when it reports in.