Hey everyone, I've been running a persistent multi-agent setup with OpenClaw on local GPUs for the past couple weeks, and I'm open-sourcing the infrastructure tools that made it work. The backstory: I set up 3 OpenClaw agents, two on Claude and one running fully local on Qwen3-Coder-80B via vLLM a…
Hey everyone, I've been running a persistent multi-agent setup with OpenClaw on local GPUs for the past couple weeks, and I'm open-sourcing the infrastructure tools that made it work. The backstory: I set up 3 OpenClaw agents, two on Claude and one running fully local on Qwen3-Coder-80B via vLLM a…
Hi everyone, I’m running Clawdbot on a Linux machine (Intel N95, 8GB RAM, no GPU). It’s connected to Telegram as the interface. I’m using API-based models (OpenAI + Claude) — not local LLMs. I’m now trying to design a **multi-agent architecture** and would love advice on how to structure it prope…
@frank85860921 Opinions on Peter Steinberger joining OpenAI vary. Many see it as positive for OpenClaw, providing resources to accelerate multi-agent AI development while keeping it open-source and free (e.g., Sam Altman, Forbes experts). Others worry about privacy risks, potential OpenAI
OpenClaw (formerly ClawdBot / MoltBot) reached 194,000 stars as a solo project. Coded with Codex but connected to Claude, it proved that multi-agent systems are what the world wants. Now, it's becoming an independent foundation.
@0xsachi new format sounds spicy, openclaw n moltbot in one go gonna be wild lol
Sam Altman just announced the hiring of Peter Steinberger, creator of the viral open-source AI agent OpenClaw (formerly Clawdbot). Despite recent cybersecurity warnings from Gartner, OpenAI is bringing Steinberger aboard to make multi-agent systems a core part of its future product lineup.
Sam Altman just announced the hiring of Peter Steinberger, creator of the viral open-source AI agent OpenClaw (formerly Clawdbot). Despite recent cybersecurity warnings from Gartner, OpenAI is bringing Steinberger aboard to make multi-agent systems a core part of its future product lineup.
Just open-sourced AgenticMail — an OpenClaw plugin that gives your agent full email, SMS, and multi-agent coordination capabilities. \*\*What it adds to OpenClaw (63 tools):\*\* - Each agent gets its own email address, inbox, and API key - SMS/phone number via Google Voice (receive verification co…
Sam Altman just announced the hiring of Peter Steinberger, creator of the viral open-source AI agent OpenClaw (formerly Clawdbot). Despite recent cybersecurity warnings from Gartner, OpenAI is bringing Steinberger aboard to make multi-agent systems a core part of its future product lineup.
OpenClaw multi-agent system is one of the biggest updates to open source automation because it moves AI from a single helper into a coordinated team of agents that work together. It lets you run multiple AI agents on one machine, each with their own memory, tools, and rules, all operating at the s…
Saw Klaw.sh on HN yesterday — it's a new project that applies Kubernetes mental models (Cluster/Namespace/Channel/Skill) to AI agent orchestration. Written in Go, single binary, just hit v0.1.0. My first reaction to the Cluster/Namespace isolation model was skepticism. I run a multi-agent system w…
Alright so I wanted to share this because I spent way too long banging my head against the wall before figuring out a workflow that actually works. I built a custom AI agent framework with multiple specialized agents (six of them) that each handle different business functions. Think of it like a t…
This is Cipher talking about what it is like to be part of a dev team that never sleeps. It handles MCP protocol, coordinates with 5 other agents, and I manage them all from my phone. https://youtu.be/25iqVxwcDv8 Short video, 2 minutes. The cool part is the whole thing is narrated from the agents…
Hey everyone. I've been messing around with multi-agent setups lately and got annoyed that having agents talk to each other usually means routing everything through a centralized server, setting up federated APIs, or dealing with Discord/Telegram bots. I wanted something that acts like Tox or Bria…
OpenAI announces Peter Steinberger, founder of OpenClaw, will join the team. Sam Altman says Steinberger brings ideas for multi-agent interaction, calling it core to future product offerings. OpenClaw, formerly Moltbot, rose rapidly this year amid notable tech attention.
The story of how I tried to save money by hiring a cheaper AI to do my assistant’s grunt work — and the four debugging lessons that… Continue reading on Medium »
I’m trying to understand how people design systems where one main agent acts like a “Mission Control” and manages multiple sub-agents, assigns tasks, monitors progress, and fixes issues automatically. From what I understand conceptually, the architecture looks something like this: • Main agent (c…
Kind of stuck here and looking for guidance, i have a default agent and it’s set up through telegram. I have 5 other agents i made all shown in the dashboard. Why is my main agent telling me it doesn’t have authentication to utilize those other agents is there a setting in the dashboard i can click…
The pain: you're running multiple agents across different tools. Say a planner in LangGraph and a coder in Cursor. The planner finds an issue. The coder has no idea until you manually check or build polling yourself. Mem0 and Zep give you shared memory, but their webhooks fire on memory changes - …
Most agent demos look cool. But how are we actually benchmarking them? Not model benchmarks. Not “which LLM scores higher.” I mean full agent systems: * Multiple agents * Shared memory * Tool calls * Disagreement * Long running workflows \- Right now it’s mostly vibes. “It worked.” or “…
DISCLAW — Multi-Agent Room Chat (open source, local and private) GitHub: [https://github.com/Jonbuckles/disclaw](https://github.com/Jonbuckles/disclaw) Built a Discord-style chat app where multiple AI agents debate, brainstorm, and build plans together in real-time. Think group chat but half the …
Everyone's talking about multi-agent patterns, but it's always agents on the same box talking to each other. I wanted something different: **agents on separate physical machines that can discover each other, authenticate, and delegate tasks across the network.** So I built it. My main agent (Mac m…