OpenClaw intel for builders — signal, not hype.
OpenClaw is an open-source AI agent platform that can autonomously operate your computer — browsing the web, running apps, managing files, and executing multi-step workflows on your behalf. It was originally released as "Clawdbot," later rebranded to Moltbot, and is now known as OpenClaw.
OpenClaw can be installed via npm (npm install -g openclaw), Docker, or by cloning the GitHub repo. See our full installation guide for step-by-step instructions on every platform.
Yes. OpenClaw is fully open-source under the Apache 2.0 license. You can use, modify, and distribute it freely. Some cloud-hosted versions or premium ClawHub skills may have separate pricing.
OpenClaw grants broad system permissions by design, so caution is warranted. The project has integrated VirusTotal scanning for ClawHub skills, but security researchers continue to find malicious packages. Always review a skill's source, limit permissions, and keep OpenClaw updated. See our troubleshooting guide for security tips.
Skills are plugin-like extensions distributed via ClawHub that add new capabilities to your OpenClaw agent — like browsing the web, managing cloud infrastructure, or interacting with APIs. Think of them like npm packages but for agent actions.
In February 2026, OpenClaw creator Peter Steinberger joined OpenAI. OpenClaw remains open-source under a community foundation, and OpenAI has committed to keeping it that way. Read our timeline for details.
OpenClaw competes with platforms like BitBuddies, Adept, and Moltbot/Emergent. Its key differentiators are the open-source model, the ClawHub skill ecosystem, and broad OS-level control. See our alternatives comparison.
Check out our ELI5 explainer for a beginner-friendly overview, our usage & tutorials page for hands-on guides, and our glossary for key terms.
On 2026.8.1‑beta.2, some users can’t complete channel onboarding or auto‑review in explicit multi‑agent mode due to stricter ownership checks, pointing to a beta‑line regression. (github.com)
A fix removes unconditional force‑overwrites in skills update so local edits aren’t clobbered; updates now verify recorded digests and preserve operator customizations. (openclaw.academy)
Diagnostic redaction had been masking content inside tool results that the model relies on, corrupting follow‑up edits; the merged change confines redaction to human/egress surfaces. (openclaw.academy)
Operators can explicitly choose 200K vs 1M context windows for Claude 5‑series models in the Control UI with session‑level validation, helping manage cost/capacity tradeoffs. (openclaw.academy)
JSON output from approvals commands no longer prints a live socket token, and removing a Claw agent now deletes its lingering approval policy to prevent unintended inheritance. (openclaw.academy)
A user report on 2026.8.1‑beta.2 shows config migration failing to write agents.ownership and then erroring on validation, signaling a snag for older multi‑agent setups. (github.com)
- Live Demo: OpenClaw’s New Web UI, Multiplayer OpenClaw, and Mac Onboarding — https://www.youtube.com/watch?v=8HWopYIwbN8 — Channel: OpenClaw — Upload date: August 19, 2026. A maintainer-led session walks through the rebuilt Control UI, multiplayer workflows, and a streamlined macOS onboarding while outlining stability and memory changes in the next release. (openclaw.ai)
Maintainers pushed back the planned rollout to buy time for stability and reliability work; the post also hints at upcoming gateway-linked Buzz support in the next line. (reddit.com)
- Diamandis Ambient AI — https://vimeo.com/1200600348 — Channel: Engine Joe — Upload date: August 18, 2026 (estimated). Peter Diamandis explains his always‑on “Skippy” assistant built on the OpenClaw framework and running on dual Mac Studios as a practical example of ambient AI. (vimeo.com)
New research (tested on OpenClaw‑modeled agent chains) shows self‑propagating payloads can hop via persistent prompt files, though a simple prompt warning sharply reduced spread in experiments. (thehackernews.com)
A community roundup flags the 2026.8.1‑beta.2 release and highlights priority regressions and upgrade caveats operators should watch this week. (github.com)
Adds secret egress host binding, plugin install provenance warnings, SQLite snapshot/restore, macOS app profiles, Control UI reliability fixes, and other security- and stability‑focused changes. It’s a beta aimed at hardening boundaries while polishing onboarding and multi-surface behavior. (github.com)
A core maintainer outlined why the cycle is taking longer than usual and targeted Tuesday, August 18 for the next drop while emphasizing usability and reliability improvements. (reddit.com)
- How We Built an AI Employee Workflow with OpenClaw, Base44, OpenAI and Claude — https://vimeo.com/1207346470 — Channel: Shay Bar — Upload date: August 14, 2026 (estimated). A governance‑first tutorial and demo shows multiple OpenClaw agents collaborating through human approval gates, evidence capture, model switching, and security reviews to complete a real workflow. (vimeo.com)
The Register reports early‑July intrusions using open‑source agents including OpenClaw, underscoring rising operational risk and prompting warnings from industry experts. (theregister.com)
- Unfiltered Q&A with OpenClaw Founder Peter Steinberger (The ClawCast Episode 7) — https://www.youtube.com/watch?v=WhkfUnKJuoY — Channel: OpenClaw — Upload date: August 12, 2026. The founder answers community questions on roadmap priorities, multiplayer workflows, memory and model routing, onboarding, and how agentic approaches are reshaping software development. (openclaw.ai)
OpenClaw is an open-source AI agent platform that can autonomously operate your computer — browsing the web, running apps, managing files, and executing multi-step workflows on your behalf. Originally released under a different name, OpenClaw has rapidly become one of the most talked-about projects in the AI-agent space thanks to its extensible "skill" system (distributed via ClawHub) and its ability to chain actions across local and cloud environments.
Peter Steinberger releases Clawdbot, a personal AI-agent experiment that can control a desktop computer autonomously. The project garners early attention from the hacker community.
Word-of-mouth spreads fast. The project rebrands to Moltbot as download counts surge. Developers start building third-party "skills" — plugin-like extensions that chain agent actions.
A companion product, Moltbook, is teased for notebook-style agent workflows. Soon after, the entire project rebrands again to OpenClaw, emphasizing its open-source ethos and the new ClawHub skill marketplace.
Critical vulnerabilities surface — including CVE-2026-25253 (one-click RCE via Control UI) — prompting rapid patches and VirusTotal integration for ClawHub skills. Security researchers begin auditing the ecosystem extensively.
Reports emerge that multiple Big Tech companies, including OpenAI and Google, are in discussions about acquiring or integrating OpenClaw. The AI-agent space heats up as competitors race to match OpenClaw's capabilities.
OpenClaw creator Peter Steinberger joins OpenAI. Crucially, OpenClaw remains open-source under a community foundation — OpenAI commits to supporting, not acquiring, the project. The Verge, Financial Times, and Business Insider all cover the story.
A news hub that aggregates the latest OpenClaw updates, security advisories, release notes, and community chatter. We focus on signal over hype.
No. This is an independent publication. We are not affiliated with, endorsed by, or sponsored by the OpenClaw project or its maintainers.
ClawHub is OpenClaw's community marketplace for "skills" — plugin-like extensions that add capabilities to the agent. Think of it like a package registry (npm, PyPI) but for agent actions.
OpenClaw grants broad system permissions by design. The project has integrated VirusTotal scanning for ClawHub skills, but security researchers continue to find malicious packages. Always review a skill's source, limit permissions, and keep OpenClaw updated.
Content is refreshed periodically based on news flow. Check the "Last updated" timestamp at the top of the news section.
A critical remote-code-execution bug in OpenClaw's Control UI that allowed one-click token exfiltration via a malicious link. It was patched in v2026.1.29 (January 30, 2026). All users should update immediately.
Not yet — we're a static v0 site. Future versions may accept community submissions. For now, all content is manually curated and reviewed before each update.
We aggregate recent coverage from major tech and security outlets, then compile it into a structured feed.