OpenClaw, the popular open-source personal agentic artificial intelligence framework, today announced the launch of version 2.0, the largest update in the agent’s history.

Unlike a traditional chatbot like ChatGPT, which lives in a browser tab and only answers questions, OpenClaw acts like a digital secretary that can actually execute tasks on a user’s behalf and lives on their computer or in the cloud. It is a free, open-source AI agent that many users have turned into a coding agent or personal assistant to organize their inbox, manage their calendar and check in on their everyday activity.

Although the version release is 2026.8.1, the OpenClaw team is calling this 2.0 because it is a massive update including mostly quality-of-life updates, better persistence, easier installation and increased security.

Up front, the OpenClaw team streamlined the installation experience so that new and veteran users can detect what’s already on their personal computer or cloud instance. For example, active ChatGPT or Claude subscriptions, application programming interface credentials and locally hosted models. The team said certain configuration options are set to defaults so users can start talking to their agent sooner, instead of getting bogged down guessing what’s available.

The dashboard has also been updated to become a real control center. Chat is now front and center and looks immediately like other AI interfaces users expect. With a text box that leads into a conversation. It allows users to start a new chat, rejoin ongoing work and watch an AI agent doing live work.

OpenClaw has also rebuilt the interface to be more interactive with new widgets and dashboards. Users can bring widgets into chats, pin them to dashboards and assign actions that can be triggered by events.

In particular, the team said it removed one previous limitation. Groups couldn’t work together with the same AI agent. Now, shared cloud sessions change that: team members can join ongoing sessions and continue work where it left off, assign tasks to teammates with full context, and collaborate in real time on content and automations.

The OpenClaw team said it now uses this “multiplayer” collaborative feature to build the AI agent framework code itself.

Other quality-of-life updates include an updated search that lets users look through past conversations using exact text, words, or phrases, and reopen the surrounding messages. This will let users easily find past work and resume where they left off.

Security updates include explicit trust boundaries, least-privilege controls and stronger protection against prompt abuse. OpenClaw is now aligned around who can trigger it, what it can access, where code executes and where credentials can be sent. It does this by controlling the gateway, providing permissions based on role and tool handling.

Early OpenClaw versions and installations suffered from major security flaws, so these updates help reduce the attack surface. Users are still expected to harden their installation against the open internet and maintain strong authentication if they choose to connect personal information or use tools that can execute on their cloud service or local machine.

This is the first major version update to OpenClaw since OpenAI Group PBC hired Peter Steinberger, the creator of the AI agent framework.

Image: OpenClaw