What Is A2A? How AI Agents Talk to Each Other
A2A, or Agent2Agent, is an open standard that lets AI agents from different makers find each other, share tasks and return results. Here is how it works.

How AI agents are built, evaluated and deployed. Recent posts compare multi-agent orchestration frameworks like LangGraph, AutoGen and CrewAI, agent sandboxes such as E2B and Modal, and agent benchmarks including SWE-bench, GAIA and OSWorld.
A2A, or Agent2Agent, is an open standard that lets AI agents from different makers find each other, share tasks and return results. Here is how it works.

Agent memory lets AI agents carry notes, preferences and decisions into later sessions. How it works in Claude Code, Codex and GitHub Copilot.

An agent harness is the software around an AI model that runs it in a loop, gives it tools and manages what it sees. Here is what each part does.

AI agents follow instructions hidden in what they read. OWASP says it is unclear if fool-proof prevention exists. My view: limit what a hijacked agent can do.

Subagents are helpers an AI agent hands a side task to. Here is how they work in Claude Code, Cursor, Codex and Copilot, and what they cost in tokens.

A cloud coding agent writes and tests code on a remote computer, then hands back changes for review. How Cursor, Codex, Copilot and Claude Code do it.

LangChain raised $300 million at a $3 billion valuation in a round led by Alphabet's CapitalG, The Wall Street Journal reported on November 12, 2026.

Databricks made Genie in Slack generally available. Teams can @mention it for answers and charts from company data, schedule analyses, and approve MCP actions.

VS Code 1.146 deprecates the Local agent harness, with removal set for February 2027, and adds an agent inbox, chat windows and Skills over MCP.

SteerDistill has a BF16 teacher coach a 4-bit agent through its mistakes, then trains the fixes in. Average score lands within 1 point of BF16.

NVIDIA gave OpenShell, its open source runtime that sandboxes AI agents and enforces what they can touch, to the CNCF as a Sandbox project at KubeCon.

3CX V20 Update 10 lets ChatGPT or Claude query call data over MCP, adds Grok for AI agents, a backup route when the AI fails, and transcription on a Mac.

Mate is a free AI shopping agent you text on iMessage or WhatsApp. It finds codes and cashback at 200,000+ US stores, checks out and tracks packages.

Helsing says its Clarity platform runs hundreds of AI agents to find military assets in battlefield data, and an operator authorises every hand-off to a weapon.

At Baidu World 2026, Robin Li launched Ciyuansuo, a marketplace that checks and rates sellers of AI model tokens. How it works and what else Baidu showed.

Gemini API managed agents can now run shell and file tools in a self-hosted sandbox over outbound HTTPS. Here is how it connects and its preview limits.

Continual-ARC, a 10 Nov 2026 preprint, finds agent harnesses as shipped save little on returning ARC puzzles, while weight training also helps on new ones.

Lumbra, whose agents it says run on top-secret Pentagon networks, raised a $25M Series A co-led by Point72 Ventures and Ansa to move into regulated sectors.

KAIST's WikiTrace links each agent memory lesson to its source run, lifting average score to 90.2 from 79.5 and cutting memory-induced failures.

Dylan Patel says persistent agents (Perplexity Computer and Claude Tag) went from 0% to 45% of SemiAnalysis AI spend. What they are and what he says.
