OpenClaw Introduces Multiplayer for Shared AI Work

OpenClaw announced Multiplayer, a free, open-source feature that lets multiple people work with the same AI agent in a shared workspace. The OpenClaw Foundation says it is designed to reduce lost context and conflicting instructions when teams use separate agents. The feature was demonstrated at GitHub Universe in San Francisco, where attendees watched two laptops connect to one agent and workspace.

In the demonstration, a participant typed “Add the Statue of Liberty” on one laptop, and the prompt appeared live on the second. After the agent added a low-poly Statue of Liberty to a 3D rendering of Fort Mason, a second participant entered “Make that three times bigger” from the other laptop. The agent enlarged the same object without requiring the participants to restate the earlier prompt or explain which object they meant. OpenClaw says Multiplayer sends requests through a shared queue, letting participants see when the agent is busy.

The Foundation presents shared context as a way for colleagues in Engineering, Product, QA, DevOps, and other departments to contribute to the same agent instead of relaying summaries between separate agents. It says Multiplayer does not fix poor management, but addresses friction caused by people having to reconstruct conversations and instructions. The Foundation reports that more than a dozen full-time maintainers work with a substantial contributor community, including AI-native developers producing large volumes of pull requests; it says keeping them aligned can be difficult.

The Foundation says around 50+ people have already collaborated in one shared environment, and it is continuing to develop the feature. It also cites Focus Week initiatives, in which maintainers and contributors coordinate around specific project priorities. At GitHub Universe, the team counted 11 spontaneous “Oh my god” reactions during demonstrations, alongside attendees’ comments such as “Wow.”

OpenClaw is developed by a nonprofit open-source foundation and is MIT licensed, model-agnostic, and customizable. Users supply their own model, including OpenAI, Qwen, Claude, or a local model, and can run OpenClaw on their own machine, a VPS, or another preferred location. The Foundation says the project has 83,000 forks and invites users to modify and extend the software; Multiplayer is available for free.

OpenClaw announced Multiplayer, a free, open-source feature that lets multiple people work with the same AI agent in a shared workspace. The OpenClaw Foundation says it is designed to reduce lost context and conflicting instructions when teams use separate agents. The feature was demonstrated at GitHub Universe in San Francisco, where attendees watched two laptops connect to one agent and workspace.

In the demonstration, a participant typed “Add the Statue of Liberty” on one laptop, and the prompt appeared live on the second. After the agent added a low-poly Statue of Liberty to a 3D rendering of Fort Mason, a second participant entered “Make that three times bigger” from the other laptop. The agent enlarged the same object without requiring the participants to restate the earlier prompt or explain which object they meant. OpenClaw says Multiplayer sends requests through a shared queue, letting participants see when the agent is busy.

The Foundation presents shared context as a way for colleagues in Engineering, Product, QA, DevOps, and other departments to contribute to the same agent instead of relaying summaries between separate agents. It says Multiplayer does not fix poor management, but addresses friction caused by people having to reconstruct conversations and instructions. The Foundation reports that more than a dozen full-time maintainers work with a substantial contributor community, including AI-native developers producing large volumes of pull requests; it says keeping them aligned can be difficult.

The Foundation says around 50+ people have already collaborated in one shared environment, and it is continuing to develop the feature. It also cites Focus Week initiatives, in which maintainers and contributors coordinate around specific project priorities. At GitHub Universe, the team counted 11 spontaneous “Oh my god” reactions during demonstrations, alongside attendees’ comments such as “Wow.”

OpenClaw is developed by a nonprofit open-source foundation and is MIT licensed, model-agnostic, and customizable. Users supply their own model, including OpenAI, Qwen, Claude, or a local model, and can run OpenClaw on their own machine, a VPS, or another preferred location. The Foundation says the project has 83,000 forks and invites users to modify and extend the software; Multiplayer is available for free.