OpenClaw 2.0 landed on August 30 as the project’s biggest release so far. It draws on work from 933 contributors and more than 16,000 pull requests—about half of every code submission the project has ever received. Among those contributors, 569 were making their first contribution.
The release started with a narrower goal: simplify installation and rebuild the browser application. It expanded into a full infrastructure overhaul. Before this version, the team had shipped 106 updates across 230 days, most of them arriving within a day or two of one another. The nearly seven-week pause before 2.0 stood out as unusual. Development itself had sped up as the contributor base grew, moving faster than the platform’s foundation and existing shipping process could handle.
The update reaches installation, messaging, memory, skills, models, automations, the browser and native apps, plugins, and security.
Streamlined Installation That Starts with What You Already Have
The new setup flow looks first at software already present on a user’s machine. OpenClaw 2.0 can detect existing ChatGPT or Claude subscriptions, API keys, and locally installed models such as those from Ollama and LM Studio. Large parts of the earlier configuration steps have been removed or postponed. Users can reach their first conversation more quickly and finish remaining setup by talking directly with the assistant.
Redesigned Control Dashboard and Faster Load Times
The browser-based Control dashboard has been rebuilt from the ground up. Project benchmarks show it now loads in roughly 575 milliseconds, down from about 1.6 seconds. New docked panels give users a file editor, a git-backed changes panel that surfaces pull-request status and CI summaries, a browser tool for element inspection and screenshot annotations, and a full-screen web terminal.
Sessions and transcripts no longer live in flat files. They have moved to SQLite. The team notes this change is not easily reversible and advises creating a verified backup before upgrading.
Shared Cloud Sessions for Team Collaboration
One of the more prominent additions is shared cloud sessions. Team members can join or take over an AI-assisted task without losing the context already built up. The OpenClaw team reports that it now uses the feature internally while developing OpenClaw itself.
The project describes itself as suitable for single-operator and single-team deployments. Shared sessions are presented as convenience tools rather than mechanisms for security or tenant isolation. OpenClaw continues to ship under the MIT license. The team has stated: “We are not selling anything here or asking you to trust one company, one model, or one AI provider.”