Is multiplayer AI the same as pair programming with an AI?

No. Pair programming puts one person and one model in an editor. Multiplayer AI puts the agent somewhere shared, where several people and automations can all work with the same run.

Can two people message the agent at once?

Yes. A message sent while the agent is working either steers the current run or queues behind it, and every message carries its author, so the agent knows who asked for what.

Who can see and steer a thread?

Every member of the workspace can see its threads. Owners, admins and members can send messages, approve tool calls and stop runs; viewers get read-only access.

Does a multiplayer coding agent need a cloud sandbox?

Mostly, yes. A run tied to one laptop is hard to share or hand off. Hoplite gives each cloud thread its own workspace cloned from your repository, with your setup script and encrypted environment variables. Threads that run locally in Hoplite Desktop can still be shared with the team.

Which models can the team use?

Models from Anthropic and OpenAI, plus others through OpenRouter. The model can be switched per message, and everyone in the thread works with the same conversation.

Can automations join a thread too?

Yes. Scheduled, GitHub, Sentry, Linear, Slack and webhook automations can start threads or post into a pinned one, so people and systems end up working with the same agent.

Try Pro free for 14 days— full access, cancel anytime

Multiplayer AI

The multiplayer coding agent for your whole team

One agent, one thread, and the whole team in it. Start a task, watch it run, jump in with a correction, answer the approval. The work lives in a cloud sandbox, not on one person's laptop.

Start freeRead the docs

Definition

What Is A Multiplayer Coding Agent?

A multiplayer coding agent is a cloud coding agent that more than one person can drive. It runs in a sandbox instead of on someone's machine, so the thread, the diff and the preview it is checking are visible to the whole team as they happen.

Multiplayer AI is the wider version of that: one agent context that teammates, integrations and automations can all write into. In Hoplite every cloud thread works this way by default. Open it, add a message, answer an approval or pick up the review. No invite needed.

Shared threads

One Thread, Everyone In It

Start a task and it shows up under Threads for the whole workspace, not only in your tab. Tool calls, diffs and test output stream to whoever has it open, and the header shows who is viewing or writing a message.

Hand off mid-task

Pick Up Where A Teammate Left Off

Someone signs off at 6pm and the agent keeps working in its cloud sandbox. Whoever is around next can reply in the same thread and change course. Each thread keeps its own workspace, restored from a snapshot on the same branch when the conversation picks back up.

Approve together

Sensitive Actions, Approved By Anyone

When a tool call needs approval, the run pauses in the thread. Any owner, admin or member of the workspace can approve or reject it, so the run isn't stuck waiting for the person who started it.

Share a local agent

Turn A Local Agent Into A Team Agent

Hoplite Desktop runs threads on your Mac with Claude Code or Codex. Share one and it becomes a project the team can open: teammates can send messages, approve tools and stop the run, and each action is recorded with their name. If the Mac goes offline, their messages queue until it reconnects.

Comparison

Solo Agent Vs. Multiplayer Agent

A solo coding agent on a laptop compared with Hoplite
Question	Solo agent on a laptop	Hoplite multiplayer agent
Who can see the run	One developer	The whole workspace, live
Who can steer it	Whoever owns the terminal	Anyone in the thread, and the agent knows who said what
When you log off	The session ends with you	Cloud threads keep running in their sandbox
Approving sensitive actions	Blocked until that one person is back	Any owner, admin or member
Parallel tasks	Port conflicts in one checkout	A sandbox per thread
Verification evidence	Scrollback, if it was saved	Tests, previews and diffs in the thread

FAQ

Multiplayer AI, Answered

Is multiplayer AI the same as pair programming with an AI?: No. Pair programming puts one person and one model in an editor. Multiplayer AI puts the agent somewhere shared, where several people and automations can all work with the same run.
Can two people message the agent at once?: Yes. A message sent while the agent is working either steers the current run or queues behind it, and every message carries its author, so the agent knows who asked for what.
Who can see and steer a thread?: Every member of the workspace can see its threads. Owners, admins and members can send messages, approve tool calls and stop runs; viewers get read-only access.
Does a multiplayer coding agent need a cloud sandbox?: Mostly, yes. A run tied to one laptop is hard to share or hand off. Hoplite gives each cloud thread its own workspace cloned from your repository, with your setup script and encrypted environment variables. Threads that run locally in Hoplite Desktop can still be shared with the team.
Which models can the team use?: Models from Anthropic and OpenAI, plus others through OpenRouter. The model can be switched per message, and everyone in the thread works with the same conversation.
Can automations join a thread too?: Yes. Scheduled, GitHub, Sentry, Linear, Slack and webhook automations can start threads or post into a pinned one, so people and systems end up working with the same agent.

Give your team a multiplayer coding agent today

Connect a GitHub repository, invite your team and start a thread. Try Pro free for 14 days.

Start freeRead the docs