OpenClaw Ships Its Biggest Update Into a Highly Crowded Field of Competitors
Trending Topics Jakob Steinschaden ● Covered by 6 sources
OpenClaw 2.0 just landed, with its biggest update yet. It adds cloud-run sessions, tighter security, and a rebuilt chat app — right as rivals crowd in.
Based on reporting by Trending Topics, Jakob Steinschaden — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenClaw has shipped the biggest update in its history, and the scale alone tells part of the story. Version 2.0 bundles more than 16,000 pull requests from 933 contributors, including 569 first-timers. The team says that is about half of everything ever merged into the project.
What began as a push to simplify installation and redo the browser interface turned into a full rebuild. The release touches installation, messaging, memory, skills, models, automations, browser and native apps, plugins and security. The new setup process now inspects what is already on the machine — subscriptions, API keys, local models — and trims away some early configuration so people can get to the first conversation faster.
The browser app has also been reorganized around chat, with a sidebar, docked panels for files, Git changes and the built-in browser, streaming Markdown, dictation, queued follow-up reordering and a side-thread mode for quick detours. Underneath, session transcripts now live in SQLite instead of files, which speeds up history loading but means downgrades need a backup. That same database setup enables full-text search across older conversations, while MCP servers can now be managed from Settings.
The bigger architectural change is that sessions no longer have to stay local. OpenClaw 2.0 can run work on paired devices or cloud workers, carrying the session workspace with it. The project also adds a durable progress card that survives reloads, structured questions that can be answered in cards, buttons or plain text, and interactive widgets that can be pinned to dashboards. On the security side, credentials can be requested through masked prompts, automations now need permission for one exact operation, and mobile and network access get stricter controls.
The timing matters because OpenClaw has gone seven weeks without a release after previously shipping 106 versions in 230 days. Hannes Rudolph says the team had grown faster than both the technical base and the shipping process, so both had to be reworked together. That pause also gave them time to make fresh installs and upgrades safer, which is the sort of unglamorous work that usually decides whether a major rewrite survives contact with users.
All of this lands in a field that is suddenly packed. Hermes Agent, LangGraph, AutoGen, CrewAI, Paperclip, Google’s Remy effort, Meta’s Hatch, Microsoft’s Scout and OpenAI’s ChatGPT Work all crowd the same broad space in different ways. OpenClaw still leans personal and local, but with cloud sessions now in the mix, the old line between a private assistant and a managed service is getting thinner by the month.
My take — AI-written commentary, not fact-checked reporting
This is the awkward truth for personal AI agents: the feature list is easy, the trust model is the product. OpenClaw keeps adding power, but the more it reaches into messages, files and cloud workers, the more it starts looking like every other system that promises convenience and then asks for a little more permission. That’s not innovation, that’s the usual software tax in a nicer jacket.
Read more about this at: Trending Topics
