OpenClaw has released the biggest update to the self-hosted open-source AI agent platform since its launch. As per the company, the update was developed by 933 contributors and includes more than 16,000 pull requests. Dubbed OpenClaw 2.0, it brings changes across installation, messaging, memory, skills, models, automations, browser and native apps, plugins and security. OpenClaw has rebuilt its browser app as a first-class experience, while the initial setup process for new users has been simplified, too.
OpenClaw 2.0 Changes
In a blog post, OpenClaw said that the 2.0 update affects every part of the platform. Described as the largest release in its history, it comes after nearly seven weeks of development. This is said to be a significant departure from the project’s earlier release cadence. The company previously shipped 106 releases in 230 days, with most updates arriving within a two-day window.
Among the most notable changes is the installation process. According to the company, new users can start with resources already available on their computers, including existing ChatGPT or Claude subscriptions, API keys, and local models. To further simplify things, some of the configurations have been moved out of the initial setup. This reportedly allows users to get to their first conversation faster and complete additional configuration later on through conversations.
OpenClaw has also rebuilt the browser as a first-class experience rather than an afterthought. Through this, users can continue setting up their Claw, return to ongoing work and follow tasks in real time.
Another addition is the ability to search past conversations. The company says users can search visible conversation text using exact words or phrases and reopen the surrounding messages from a matching result. Meanwhile, support for sessions beyond the Gateway allows them to continue work on paired devices or cloud workers, move a session’s workspace with it and reuse machines and project seeds for subsequent cloud sessions.
With new durable session progress cards, users can also follow their work in real time. This includes subagent activity and accumulating edits after reloading the web or native app. There are now widgets and dashboards that make up a more interactive interface. Users can add widgets inside chats, pin them to session dashboards, grant specific actions or network origins and export rendered views as images.
For recurring tasks, users can now approve an exact operation once and allow an automation to perform it again. The update also expands media support and adds new native playback controls. As part of the improved security infrastructure, OpenClaw 2.0 will allow an agent to request credentials through a masked prompt without exposing the value in the chat or model context.
Despite the extensive list of changes, OpenClaw said that the platform remains self-hosted and open-source. Developers will be able to build on it without being tied to a single AI model or provider.
