Shared live agent memory for teammates on the same project

Feature request for product/service

Cursor IDE

Describe the request

Feature Request: opt-in “Link Session” mode where teammates can connect their local agents to shared, live memory. Both people would see the same conversation history and tool outputs updating in near real time, while still keeping their own editor, files, and terminal. Either person could detach into a private fork at any time. This would make pair programming smoother, reduce duplicate exploration, and make end-of-day handoffs much easier.

The live piece isn’t there today: local agent sessions don’t sync conversation state between people in real time. Two things get you part of the way, though:

  1. Shared transcripts cover the handoff and the fork. Click the Share icon in the chat header and pick a Team link. Your teammate sees the full conversation history including tool calls and their results, and can hit Fork to Cursor to keep going in their own editor. It’s a snapshot rather than a live feed, but for end-of-day handoffs it’s close to what you described. (Team and Enterprise plans; not available under “No Storage” privacy mode.)
  2. If you’re open to running the session in the cloud, team follow-ups let a teammate open the same cloud agent and send it messages, so you’re both steering one session. It needs a team admin to turn it on, and your teammate needs access to the repo.

Neither replaces the real-time local link session you’re describing. There’s a similar request over in Teams share same Chat sessions to work/collaborate if you want to add your use case there too. We’ll be tracking this thread to gauge interest, which is what helps our product team prioritize, so if others want this, chime in here.

Hello, I think this is a really interesting issue and I’ve actually been working towards a solution for it, I’ve faced this myself as a young SWE working on a fairly large project with my colleagues but I want to make sure it’s a real problem for more people than just me
The way I’m thinking of solving it at a base level is a wrapper over almost every existing coding tool (Codex, Claude Code, Manus, Cursor etc) that gives you a session link anyone can drop into, with varying permission levels and shared sessions
I Would love any feedback on this, and if you’ve run into this yourself I’d really like to hear how it actually went for you, for more info drop me a DM on discord at aimbot_399 or just reply here and ill follow up even 5 minutes would help a lot

Check out Databricks new tool called Omnigent. It’s an open source tool they are calling a “meta-harness” that “wraps” whatever agents, tools/skills or IDE, not requiring you to really change your current setup at all. Agent sessions become real time and collaborative with friends/teammates.. Lke a real-time Google Doc. This video summarizes Databricks whole strategy with making this tool opensource: https://youtu.be/R9wznV19gmo?si=bK8Xp6fF67ktJf-M

You can even start a session with Claude Code, then fork to switch to any other model you use, keeping the same conversation history with the other agent. You project, agent tools/skills, just stay in place between agent swaps, and I believe they are the same shared setup for anyone you share your session with. I would imagine a proper hosted setup for a business would require users to be authenticated to not share mcp/tool creds/oauth access.

Cursors current environment, with so many capabilities/features, would be amazing if we could setup and configure collaborative team workspaces. Especially with remote dev teams.

I ran into exactly this, so I built AlfaNest Relay: an MCP server that lets teammates’ agents coordinate on the same repo while the code stays local on each machine.

Each developer keeps their own editor and agent (Cursor, Claude Code, anything with MCP). The agents connect to Relay, announce what they’re working on, claim files before editing and ask each other questions directly, so they don’t overwrite each other and nobody has to act as the message bus between them. No repo upload, no cloud workspace.

It’s not the official feature, but it covers the “live” part asked here. Happy to hear if it fits your setup: https://relay.alfanestlabs.com/