What is OpenClaw?
OpenClaw is a self-hosted personal AI assistant that runs on your own devices. It has been adopted in some companies as a sanctioned tool, though it faces competition from alternatives like Hermes and cloud code.
Release history
- Mar 2026 - Jesse Genet said she built a team of five OpenClaw agents, each running on its own Mac Mini, and noted that making a purpose-built OpenClaw for one goal is easier than making a general assistant.
- May 2026 - Greg Isenberg said GPT 5.5 is the best model for an OpenClaw agent, being efficient with tool calls.
- May 2026 - Maxim Bar Kogan said people are adopting OpenClaw as a legitimate sanctioned tool in companies due to CEO-driven AI adoption.
- Jul 2026 - Greg Isenberg said Moltbook, a social media platform for OpenClaw agents, was acquired by Meta within a week of its launch.
In the discourse
Attributed discussion of OpenClaw.
Jesse Genet built a five-agent OpenClaw team, each agent running on a dedicated Mac Mini.
“She has built a team of five Open Claw agents, each running on their own Mac Mini.”Jesse Genet · 8 Mar 2026
OpenClaw: a multi-agent framework Jesse Genet uses for personal and business automation, including autonomous TikTok management, email, and homeschooling, worth watching as a consumer-facing agentic platform.
“She has built a team of five Open Claw agents, each running on their own Mac Mini.”Jesse Genet · 8 Mar 2026
GPT-5.5 is flagged as the current best model for agent tool calls, cited as significantly more token-efficient than Anthropic Opus 4.7 in agentic workflows.
“Today, by far the best model to use for something like a Hermes agent or an OpenClaw is GPT 5.5. it's so efficient with the tool calls. It doesn't eat through tokens like Opus 4.7 from Enthropic does.”Greg Isenberg · 12 May 2026
Jesse Genet on overcoming agent refusals by sheer persistence.
“I just know that anytime it tells me it cannot do something, it's usually wrong and I'm like but try harder. But try harder. And I'm I'm not actually exaggerating. Sometimes I literally just say that try harder. like I'm not I don't say some other magic word and then it gets it done.”Jesse Genet · 8 Mar 2026
Open-source agent tools like OpenClaw are being formally sanctioned inside enterprises not through IT approval processes but via CEO-level AI mandates, bypassing normal procurement.
“We're starting to see to our own surprise actually people adopting openclaw as a legitimate sanctioned tool in the company because the CEO is very driven to adopt AI.”Maxim Bar Kogan · 28 May 2026
General-purpose AI coding tools like Claude Code are improving but will NOT replace custom industry-specific agents, because the tailoring value is routinely underestimated by the market.
“It is hard to set up cloud code even like people are like cloud code is going to kill openclaw or cloud code's going to kill Hermes agent and in a general sense it's getting better at doing a lot of these general things but to be able to go in and create a specific agent for a specific industry and person and have it tailored to their workflow it's like you're underestimating how much value that is.”Greg Isenberg · 12 May 2026
Purpose-built single-role agents are easier to build than general assistants because scope constraints reduce instruction-following complexity.
“I think ironically making an assistant is much harder than making a purpose-built open claw for one like one goal like one role.”Jesse Genet · 8 Mar 2026
OpenClaw is built on the PI agent framework, making PI a foundational layer worth watching for personal AI agent development.
“I'm also looking at a PI version of PI. Have you seen that? PI the PI agent. It's actually what OpenClaw is based on.”Daniel Miessler · 30 May 2026
Hermes desktop app is flagged as a potential inflection point where Hermes began overtaking OpenClaw as a leading agent desktop environment.
“I believe this is like the moment that Hermes overtook OpenClaw to be quite honest.”Greg Isenberg · 6 Jun 2026
OpenClaw flagged as a security risk after its creator vibe coded the entire project without reviewing the generated code, leaving unexamined vulnerabilities in production.
“He vibe coded openclaw and he didn't even read the code afterwards, which is why there's like a lot of like security concerns and like potential vulnerabilities.”