Influenced by some YouTube videos, and threads here about giving cursor some long term memory, and autonomous workflow, I started working with linear.app to mange my work (and my dedicated and patient “ai assistant”). I open a ticket in linear for a new feature. I describe the high level of it, along with some requirements, and points of emphasis, and using cursor rules, and an mcp server connecting to linear (I used this one, but hope to find a better one: GitHub - ibraheem4/linear-mcp: Enables AI agents to manage issues, projects, and teams on the Linear platform programmatically.).
so now, once I have a ticket ready, I label it with cursor or add [cursor] to its title, and in cursor i type: NLF and bam… (NLF = Next Linear Feature).
Cursor will create a detailed plan and update the ticket with it, wait for my review, and once approved will create a feature branch with linear conventions, which once I merge back from a merge request will mark the ticket done.
this helps keep a nice trail of work.
any thoughts, feedback or advice is welcome. If there is interest, I can share the rules I use.
Hey @srool - Have you tried hooking up the newly released official Linear MCP to cursor yet?
I’ve tried using it and for someone reason it doesn’t work, hard to debug there isn’t much to go off of - what you tried?
2025-05-14 15:01:40.297 [info] mote: Starting new stdio process with command: npx -y mcp-remote https://mcp.linear.app/sse
2025-05-14 15:01:42.695 [info] mote: Client closed for command
2025-05-14 15:01:42.695 [error] mote: Error in MCP: Client closed
2025-05-14 15:01:42.696 [error] mote: Failed to reload client: MCP error -32000: Connection closed
I’m also having issues getting this to work. in the cursor chat, cursor can see the mcp configuration for linear but says “I don’t have direct access to make MCP calls through available tools” is there a security setting in cursor that prevents this?
I was having issues also with the old Linear MCP (open source) I used with some of the LLMs. I mostly gave up and do some of the ticket housekeeping manually 
solved it - the authentication was working. to use the official MCP I had to start my agent commands with “Linear:” to properly use the mcp. seems a bit too specific for an ai, but it works