OpenClaw 2.0 brings simplified setup, a rebuilt browser app, and multiplayer sessions

Details

The OpenClaw Foundation has shipped version 2.0 of its open-source AI platform, calling it the largest release by volume—over 16,000 pull requests went into this version covering onboarding, collaboration, compute options, and the core browser interface. The biggest user-facing improvement is simplified first-run setup. The platform now scans for existing AI resources on the machine—ChatGPT or Claude subscriptions, API keys, local models—and skips manual configuration for whatever it finds. Users can refine settings through a conversation with the assistant rather than editing config files. Collaboration is new in 2.0: Shared Cloud Sessions allow multiple users to work on the same task simultaneously, with full context passed to incoming team members. The OpenClaw team reports using this feature to build OpenClaw itself. Sessions can run locally, on paired personal hardware via openclaw connect, or on disposable rented machines via Crabbox (supporting AWS and Hetzner backends). Provider credentials stay on the user's gateway in all cases—they never reach remote machines. The rebuilt browser app adds a Session Rail—a compact in-chat widget showing ratings, plan progress, and pull requests for a running session—and a companion thread for asking questions without interrupting the active agent. The release also covers memory, skills, model support, automations, native apps, plugins, and security changes.