On November 5, 2026, the open-source agent project OpenClaw announced Multiplayer on its official blog: several people can connect to the same agent at once, sharing one context in one workspace. Until now, each department ran its own agent, and the context held by product, engineering, and marketing never met. Multiplayer targets exactly that break. The feature is free, under OpenClaw's usual MIT open-source license.
The idea is easier to grasp in use. At GitHub Universe, the OpenClaw team demoed it live: two laptops connected to the same agent, with a large screen rendering a 3D scene of Fort Mason in San Francisco. An audience member typed "Add the Statue of Liberty" on the first laptop — before they finished typing, the same keystrokes were already appearing live on the second laptop. Moments after submitting, the statue materialized in the scene. Then, from the second laptop, someone typed "Make that three times bigger," and the first laptop showed the instruction as it was typed; the model did not need to ask which object, it simply followed the shared context. Requests flow through a shared queue, so everyone can see when the agent is busy instead of firing off conflicting instructions. Inside the foundation, more than 50 people already collaborate in one shared environment.
For teams, what disappears is the repeated work of syncing context: product managers no longer explain a requirement three times, engineers no longer guess which version was agreed last Tuesday, and changes land in the same conversation. Multiplayer will not fix management — the blog says plainly it cannot cure bad management or pointless meetings — but it narrows the new collaboration gap created when every department's agent knows nothing about the others. You still bring your own model: OpenAI, Qwen, Claude, or a local model all work, running on your own machine or a VPS, and 83,000 forks show the community is already reshaping it.
The feature is live with the blog post, and individuals and teams can install it from the OpenClaw website. Existing OpenClaw users do not need to switch models or migrate data; anyone wanting a preview can watch the GitHub Universe demo. Shared-agent collaboration is still new, so permission boundaries and the cost of mistakes are rules each team will have to set for itself.
