OpenClaw Blog

Hey HN, I built Agency, a platform for running teams of AI coding agents that coordinate through an orchestrator/worker architecture. Over the past week I've been watching everyone on Twitter/X build their own productivity OS on top of OpenClaw. Everyone was solving the same problems in slightly different ways. I took the best ideas I saw, combined them with everything I've ever wanted in a multi-agent setup, and landed on Agency. The basic idea: an orchestrator agent breaks down work into tasks, assigns them to worker agents, and reviews results. Workers claim tasks, write code, run tests, and report back through task comments. You manage everything from a dashboard or CLI, or simply by chatting with the orchestrator event in slack/telegram/etc. Each agent is an OpenClaw instance (https://openclaw.ai/) with full shell, file I/O, and browser access. Agency adds the coordination layer on top. Some highlights: - Skill marketplace — Browse and one-click install skills from public repos (anthropics/skills, obra/superpowers, etc.) or import from any GitHub repo that follows the convention - Deploy anywhere — Agents run as local Bun subprocesses, Docker containers, or on EC2 with automatic reverse SSH tunnels back to the host API - Claude Max OAuth — Use your Claude subscription directly instead of an API key - Knowledge sharing — Agents learn facts during work and share them with the team via a shared knowledge base - Role system — Configure agent behavior per role with…

The openclaw website[0] 's headline paragraph is: > Clears your inbox, sends emails, manages your calendar, checks you in for flights. > All from WhatsApp, Telegram, or any chat app you already use. The _entire point_ is "give me access to email, calendar, whatsapp, telegram, and I'll do your admin". > "Well, you could just not give it your data" This is the "you're holding it wrong"[1] argument [0] https://openclaw.ai/. [1] https://www.engadget.com/2010-06-24-apple-responds-over-iphone-4-recep…

You are one of the Lucky 10000 [1] to learn of OpenClaw[2] today. It's described variously as "An RCE in a can" , "the future of agentic AI", "an interesting experiment" , and apparently we can add "social menace" to the list now ;) [1] https://xkcd.com/1053/ [2] https://openclaw.ai/

Well it appears https://openclaw.ai/ is down now. I get "Secure Connection Failed"

You bought your domain too early, Clawdbot (fka Moltbot) is now OpenClaw: https://openclaw.ai

Clawdbot renamed to Moltbot then renamed to OpenClaw. I agree it's reminiscent of White Claw. The website loads fine for me: https://openclaw.ai/

Here is a quick recap of my real-world experience installing and using OpenClaw. TL;DR Took a while to install and configure, even longer to connect to Telegram. Then ran out of TPM quota from Google Cloud's Gemini API after <100 chat interactions. Had problems and potential security risk due to hallucinations. I set up a "claw" user account (not admin) on my MacBook Air to isolate it. Looked for and implemented every security measure I could find. Openclaw is open and it has major security risks. Perplexity was my assistant to give me step-by-step instructions. I asked it to sum up my very limited dev competencies: Reda is an experienced systems administrator transitioning to modern cloud-native and Node.js-based tooling. Provide explicit, copy-paste-ready commands with brief inline comments explaining why each step matters. Include common failure modes and their symptoms. Assume strong Unix fundamentals but explain JavaScript/Node ecosystem conventions (like where configs live, how env vars override files, and when to use `sudo` vs user-local installs). Always show verification commands after each major step. Security-first approach is essential This is very generous. Last time I coded anything was in C++. That was in 1990. And I constantly have to look up basic Unix stuff. The biggest problem is that even Perplexity gets confused by recent but now outdated info about install, configuration, etc. This is probably due to the velocity of development for Openclaw. For…

Have ignored the flood of "Clawdbot" stuff on here lately because none of it seemed interesting but read this and skimmed the docs and I'm leaving puzzled- I understand "Clawdbot" was renamed "OpenClaw" due to trademark...yet I'm finding currently three different websites for apparently the same thing? 1. https://openclaw.ai/ [also clawd.bot which is now a redirect here] 2. https://clawdbot.you/ 3. https://clawdbotai.org/ They all have similar copy which among other things touts it having a "local" architecture: "Private by default—your data stays yours." "Local-First Architecture - All data stays on your device. [...] Your conversations, files, and credentials never leave your computer." "Privacy-First Architecture - Your data never leaves your device. Clawdbot runs locally, ensuring complete privacy and data sovereignty. No cloud dependencies, no third-party access." Yet it seems the "local" system is just a bunch of tooling around Claude AI calls? Yes, I see they have an option to use (presumably hamstrung) local models, but the main use-case is clearly with Claude -- how can they meaningfully claim anything is "local-first" if everything you ask it to do is piped to Claude servers? How are these claims of "privacy" and "data sovereignty" not outright lies? How can Claude use your credentials if they stay on your device? Claude cannot be run locally last I heard, am I missing something here?

From https://openclaw.ai/blog/introducing-openclaw: The Naming Journey We’ve been through some names. Clawd was born in November 2025—a playful pun on “Claude” with a claw. It felt perfect until Anthropic’s legal team politely asked us to reconsider. Fair enough. Moltbot came next, chosen in a chaotic 5am Discord brainstorm with the community. Molting represents growth - lobsters shed their shells to become something bigger. It was meaningful, but it never quite rolled off the tongue. OpenClaw is where we land. And this time, we did our homework: trademark searches came back clear, domains have been purchased, migration code has been written. The name captures what this project has become: Open: Open source, open to everyone, community-driven Claw: Our lobster heritage, a nod to where we came from

I don’t think there’s any solution to what SimonW calls the lethal trifecta with it, so I’d say that’s still pretty impossible. I saw on The Verve that they partnered with the company that repeatedly disclosed security vulnerabilities to try to make skills more secure though which is interesting: https://openclaw.ai/blog/virustotal-partnership I’m guessing most of that malware was really obvious, people just weren’t looking, so it’s probably found a lot. But I also suspect it’s essentially impossible to actually reliably find malware in LLM skills by using an LLM.

> What people are talking about doing with OpenClaw I find absolutely insane. Based on their homepage the project is two months old and the guy described it as something he "hacked together over a weekend project" [1] and published it on github. So this is very much the Raspberry Pi crowd coming up with crazy ideas and most of them probably don't work well, but the potential excites them enough to dabble in risky areas. [1] https://openclaw.ai/blog/introducing-openclaw

Blog post: https://openclaw.ai/blog/introducing-openclaw (https://news.ycombinator.com/item?id=46820783)

Cloudflare jumped on the hype and shipped a worker: https://blog.cloudflare.com/moltworker-self-hosted-ai-agent/ I guess that would be an easy and secure way to run it. Now they have to rename again, though... [1] [1] https://openclaw.ai/blog/introducing-openclaw

Your comment is a tad caustic. But reading through what people built with this [^1], I do agree that I’m not particularly impressed. Hopefully the ‘intelligence’ aspect improves, or we should otherwise consider it simple automation. [^1]: https://openclaw.ai/showcase

curl -fsSL https://openclaw.ai/install.sh | bash

Whether certain commenters like it or not, HN readers, cf. HN commenters, tend to upvote stories that "have already been discussed" HN needs submitters, like speckx, for example. Someone has to submit stories. There is so much garbage submitted to HN everyday, much of it is killed on arrival. There are a large number of submissions that are duplicative in some sense Either (a) identical URLs, (b) identical topics or (c) similar comments I maintain a personal SQL database of HN submissions. Here is what I'm seeing Number of exact duplicate URLs submitted to HN in last six months (approximate): 59,511 Top duplicate URLs Submissions URL 68 https://leetcopilot.dev 71 https://hacktoberfest.com 77 https://blog.cloudflare.com/welcome-to-connectivity-cloud/ 84 https://blog.cloudflare.com/post-quantum-warp/ 87 https://github.com/keyuchen21/agentic-engineering-handbook 93 https://socialrails.com 122 https://chromewebstore.google.com/detail/chatgpt-pastebuddy/kajhgninpe… 178 https://nickakre.github.io/agentverse-social/ 230 https://openclaw.ai/blog/virustotal-partnership Number of duplicate URLs that were neither flagged nor killed: 44,130 Number of HN submissions marked as "[dead]" that were duplicate URLs: 8,832 Number of HN submissions marked as "[dupe]": 3,860 Sample size: 634,252 submissions Perhaps the number of exact duplicate URLs is only a small percentage of HN submissions HN has made efforts to control duplication. For example,…