Key facts
- OpenClaw 2.0 features a rebuilt browser app, shared multiplayer cloud sessions, and enterprise-focused features.
- The update was built by 933 contributors, including 569 first-time contributors.
- New setup process guides users based on existing subscriptions or API keys.
- The Control UI has been redesigned to be chat-centric, similar to other major AI models.
- Enhanced security features include request-specific approvals and command permissions, though sandboxing is off by default.
- The release aims to address user frustrations and compete with rivals like Nous Research's Hermes Agent.
OpenClaw has launched its most significant update to date, version 2.0, which includes a complete rebuild of its browser application, the addition of shared multiplayer cloud sessions, and features clearly designed for enterprise adoption. This release comes after a two-month development period, during which the project saw an unprecedented number of contributions, with over 933 contributors, including 569 first-timers, submitting more than 16,000 pull requests.
The update addresses long-standing user frustrations, particularly concerning setup difficulties, which have led some users to switch to competing platforms like Nous Research's Hermes Agent. OpenClaw 2.0 introduces a streamlined setup process that checks for existing subscriptions and API keys, and it validates model functionality before proceeding. The user interface has been redesigned to be chat-centric, mirroring the experience of platforms like ChatGPT and Gemini, with integrated tools such as file management, Git status, and a browser panel.
A key new feature is the shared cloud sessions, allowing multiple users to collaborate on tasks within the same agent instance without losing context. This is intended to enhance team productivity, as suggested by founder Peter Steinberger's own team's adoption of shared setups. Security has also been enhanced with features like request-specific approvals and command permissions, though advanced sandboxing and role-enforced execution are opt-in.
The development of OpenClaw 2.0 was described as "accidental" due to scope creep. What began as a focused effort to simplify installation and rebuild the browser app expanded to touch nearly every aspect of the framework, including memory, models, plugins, messaging, and security. This was partly driven by the team's internal use of multiplayer sessions, which highlighted deficiencies in the previous single-user design.
OpenClaw remains a self-hosted, open-source framework that transforms large language models into persistent agents capable of interacting via various messaging platforms and accessing user data with permission. The project is now managed by the non-profit OpenClaw Foundation, with partners including Microsoft, GitHub, Nvidia, Atlassian, and Tencent.
In the competitive landscape, Hermes Agent has gained traction by offering features like automatic skill generation from workflows and more secure default settings. OpenClaw 2.0's new security features aim to narrow this gap, though the company emphasizes that true tenant isolation requires setting up separate "cells" or Gateways.
Existing users can update via the command-line interface, with a caution to back up sessions due to a migration to SQLite storage. New users can access guided onboarding across desktop and mobile platforms, with options for importing data from other AI services.