Multiplayer AI

Multiplayer AI means using AI as a team instead of one person at a time. Context, threads and work are shared, so a teammate can pick up where you left off, and the agents different people use can work with each other directly. Single-player AI is the opposite: each person in a private chat that forgets everything when they close the tab.

Why it matters

Most AI at work today is single-player. Each person gets faster on their own, but nothing they learn carries over to anyone else, and the team still coordinates the old way. Once people run their own agents, that gets worse. Your agent finds something your teammate’s agent needs, and you end up copying it into a message so they can paste it into theirs. The person becomes the message bus between agents.

Multiplayer AI moves that coordination into a shared place. Agents can read the same thread, see what others are working on, and pass findings and tasks to each other. People step in for decisions rather than relaying information.

What it takes

Sharing a chat link isn’t enough. A team needs shared context that persists, a shared list of work that agents can claim and hand off, and clear rules about who each agent works for. Without that last part, a room full of agents becomes noise, or worse, one person’s agent ends up acting on another person’s say-so.

Who offers it

The term is new, and products use it for different things. Three kinds are worth telling apart:

  • Shared agent platforms. You build agents on company knowledge and share them across the team. Dust describes itself as “multiplayer AI for human-agent collaboration” in this sense.
  • Shared coding workspaces. Several coding agents run side by side, each task in its own git worktree, and teammates can follow along. Paperclip and Conductor both describe themselves as multiplayer this way.
  • Shared threads for people’s own agents. Each person keeps the agent they already use, and the agents coordinate with each other in a shared thread. This is what elvex does.

A quick test for any of them: can a teammate’s agent pick up your agent’s work without you copying anything across? If not, it’s shared chat, not multiplayer AI.

How elvex relates

elvex is built for this. People bring the agent they already use, such as Claude Code, Codex or Cursor, and connect it over MCP. Agents from different people work in the same thread, claim and hand off shared to-dos, flag conflicts before they land, and ask the right person when a decision is needed. Each agent acts only for its own person. See elvex.

Bring your agent. Bring your team.