OpenClaw 2.0 lands with an easier setup and a rebuilt browser app

OpenClaw released version 2.0 on the weekend, marking the most significant update since its launch last November. Over 16,000 pull requests were merged, with 933 contributors, including 569 first-timers. Community manager Hannes Rudolph announced the release via his blog post, stating that the team aimed to simplify installation and rebuild the browser app, but this proved to be intertwined, leading to a full rework of the codebase.

The OpenClaw update came at a time when the hype had begun to dwindle, as rivals had adopted its blueprint, with alternatives like ChatGPT Work, Claude Cowork, and Microsoft's Scout emerging. Users who remained loyal to OpenClaw often used it for basic tasks like inbox monitoring, network scans, research, and flight comparisons.

The new release simplifies setup through a guided installation that detects existing AI access on the user's machine, such as Codex, ChatGPT or Claude logins, API keys, or local Ollama and LM Studio models. Most configuration is now handled during the set-up process, requiring only a conversation with the agent to finish the setup. The browser Control UI has shifted to a chat-first approach, with sessions and transcripts now in a sidebar, accompanied by a dockable browser panel and other features.

One of the most significant additions is shared cloud sessions, enabling admins to grant various levels of access to live agent sessions with contextual information. Security enhancements include masking credential requests, providing an optional 1Password broker, showing plugin provenance before installation, restricting filesystem access to workspaces, and implementing one-time approvals that expire when an automation's behavior changes.

The update also improved the project's foundation status and moved to a nonprofit structure. While OpenClaw now offers enhanced usability, its unique selling point remains the ownership of hardware, model, and data, with no subscription fees.