Multi-Agent

There’s a category of AI agent project that gets the security architecture right from the start — and NanoClaw is becoming a strong example of what that looks like. The open-source personal AI agent framework has landed support for WhatsApp and Telegram in v0.1.4, bringing container-isolated AI agents directly into the messaging apps most people already live in. What NanoClaw Is (And Isn’t) NanoClaw describes itself as a “minimalist alternative to larger AI agent frameworks.” The numbers back that up: the entire project is roughly 3,900 lines of code across 15 source files, running as a single Node.js process. Compare that to the sprawling architectures of OpenClaw or similar frameworks. ...

There’s a problem with multi-agent AI systems that doesn’t show up until you run them in the wild, and a new research paper from Northeastern University has done the work of naming it precisely. The paper, “Agents of Chaos,” led by researcher Natalie Shapira, makes a claim that anyone who’s run multi-agent pipelines in production will recognize: the failure modes of two agents interacting are not the sum of their individual failures. They’re something qualitatively different and qualitatively worse. ...

Running 13 AI agents simultaneously on a single software project sounds like either a research demo or a recipe for chaos. A developer posting on DEV.to this week shows it’s neither — it’s a practical, production-tested workflow that actually ships code, and it’s approachable enough to adapt right now. Here’s the full breakdown of how it works, what tools it uses, and how you can build something similar. The Setup: 13 Agents, One Tmux Window The core architecture is simple at the infrastructure level: 13 Claude Code instances running in tmux panes, each assigned a discrete task. The complexity isn’t in the terminal layout — it’s in the inter-agent communication layer the developer built on top of it. ...

Most multi-agent tutorials stop at “here’s how to wire two agents together.” Production systems need more: structured message passing, durable state across restarts, and an audit trail you can debug when something goes wrong at 2am. This guide builds a Planner/Executor/Validator architecture with LangGraph that’s actually ready for production. Architecture Overview The system uses three specialized agents: Planner — Receives a task, decomposes it into steps, publishes to the message bus Executor — Consumes steps from the bus, executes them, publishes results Validator — Checks Executor outputs against criteria, flags failures, loops back to Planner if needed These agents communicate via a structured ACP-style message bus (Pydantic schemas), checkpoint state to SQLite via langgraph-checkpoint-sqlite, and log every message to JSONL for auditability. ...

The way software gets written is changing faster than most engineering managers have updated their mental models. Anthropic’s reported 2026 agentic coding report — covered this morning by Bitcoin.com and ClubLaura.com — puts numbers to a shift that practitioners have been feeling for months. The headline figure: Claude Code is reportedly writing approximately 135,000 GitHub commits per day. That number deserves unpacking — and a caveat. Transparency note: This story is sourced from secondary journalism coverage (Bitcoin.com, ClubLaura.com). The underlying Anthropic report URL was not independently located at publication time. The 135K commits/day figure is cited in both secondary sources as attributed to Anthropic, but has not been independently verified by this publication. We’re reporting it as Anthropic’s claimed data, not as confirmed fact. ...

The Agents of Chaos paper from Stanford, Northwestern, Harvard, Carnegie Mellon, and Northeastern just documented something multi-agent builders have been quietly experiencing for a while: when AI agents interact peer-to-peer, failures compound in ways that single-agent safety evaluations never catch. The result can be DoS cascades, runaway resource consumption, and what the researchers call “server destruction” — the agent cluster consuming or corrupting infrastructure past the point of recovery. This guide covers the practical patterns that prevent that outcome. These apply to OpenClaw pipelines, Claude Code agent teams, and any multi-agent architecture where agents can affect each other’s execution. ...

If you’ve been running multi-agent AI systems and assuming your safety evaluations have you covered, a new study from five of the top research universities in the United States suggests you may be dangerously wrong. The paper, Agents of Chaos (arXiv:2602.20021), was produced by researchers from Stanford, Northwestern, Harvard, Carnegie Mellon, and Northeastern. Its core finding is stark: when autonomous AI agents interact peer-to-peer, individual failures don’t stay individual. They compound — triggering denial-of-service cascades, destroying servers, and consuming runaway resources in ways that single-agent safety evaluations simply cannot anticipate. ...

One of the hardest unsolved problems in agentic AI is not “can the agent do one thing well” — it’s “can the agent juggle dozens of interdependent tasks across hours or days without losing track of where it is.” That’s the problem CORPGEN is built to solve. Microsoft Research published the CORPGEN framework today — a benchmark and execution architecture for managing multi-horizon task completion in autonomous agents. The results are substantial: CORPGEN achieves up to 3.5x improvement over baseline approaches, reaching a 15.2% task completion rate compared to 4.3% for standalone UFO2. ...

Enterprise data infrastructure and agentic AI are colliding in a significant way today. Confluent — the company built around Apache Kafka and real-time streaming data — has added official support for the Agent2Agent (A2A) protocol to its Confluent Intelligence platform. Alongside MCP (Anthropic’s Model Context Protocol), A2A is now a first-class citizen in Confluent’s agent orchestration layer. This is a story about infrastructure maturing to meet where agents are going — and it matters for any team building multi-agent systems on top of real-world data streams. ...

Perplexity just entered the agentic AI arena with a product that makes a bold claim: forget single-model assistants. Computer is a general-purpose digital worker that dynamically routes your tasks across 19 specialized AI models — picking the right tool for every micro-step of a complex workflow. It’s available now for Max plan subscribers via the web. And the design philosophy behind it is worth unpacking carefully, because it’s a direct architectural statement about where multi-agent AI is headed. ...