TOPAIAgent
Updates June 22-29
GOOLGE’S A2A PROTOCOL JOINS
LINUX FOUNDATION
Google has officially donated its A2A protocol to the
Linux Foundation, marking a significant step toward
open, interoperable standards for AI agent
communication.
The announcement was made at the Open Source
Summit North America, where the Linux Foundation
introduced the new, Agent2Agent Project.
Click here to learn more
GOOGLE RELEASES GEMINI CLI: NEW
OPEN SOURCE CODING AGENT
Gemini CLI is Google's open-source AI tool for
developers, offering advanced features via natural
language commands in terminals.
Gemini is leading the CLI coding race as it has the
context window of 1 million tokens, unlike Claude
code, which only supports 200k.
Click here to learn more
GEMINI ADDS AGENT MODE IN THEIR
GEMINI CODE ASSIST
Gemini Code Assist, Google’s AI-powered coding
assistant, has introduced a powerful new "agent
mode" that significantly expands its capabilities for
developers across all plans—including free, Standard,
and Enterprise—at no additional cost.
Agent mode transforms Gemini Code Assist from a
simple prompt-response assistant into a
collaborative, multi-step reasoning agent.
Click here to learn more
GOOGLE FINALLY INTRODUCES DOPPL
IN THE US
Doppl is one of the few projects that was unveiled in
Google I/O 2025.
Doppl lets you see how any outfit might look on you by
creating personalized, animated images and videos. You
can upload a photo or screenshot of any outfit—whether
it’s from a friend, a store, or social media—and Doppl
generates a visualization of you wearing it
Click here to learn more
GOOGLE’S GEMMA 3N COMES
OUT OF PREVIEW
Google officially released Gemma 3n on June 26,
2025, following a preview phase that began in May
2025.
The model is now fully available to developers and
the public, with support across major platforms and
frameworks including Hugging Face Transformers,
[Link], Google AI Edge, Ollama, and MLX
Click here to learn more
GOOGLE RELEASES MAGENTA-
REALTIME
Magenta-Realtime is Google’s latest open-source,
open-weights AI model for real-time music
generation.
Users can control the model with both text prompts
(e.g., “play synthwave with an ’80s vibe”) and audio
prompts (e.g., “continue in the style of this sample”),
thanks to the MusicCoCa embedding module.
Available to try on Colab.
Click here to learn more
MICROSOFT’S MU - AN LLM
ENABLING AGENTS IN WINDOWS
Microsoft Mu is a lightweight, on-device language
model designed specifically for edge devices,
starting with its integration into Windows 11 Copilot+
PCs.
Mu powers the new AI agent in the Windows
Settings application, enabling users to control and
query system settings using natural language—
entirely offline, with fast and private responses
Click here to learn more
DEEPLEARNING AI’S NEW COURSE
ON HOW TO USE ACP PROTOCOL
The new course is designed to help AI builders
connect and reuse multiple agents built with different
frameworks in a single system.
You will learn directly from Sandi Besen (AI Research
Engineer & Ecosystem Lead at IBM) and Nicholas
Renotte (Head of AI Developer Advocacy at IBM).
The course is around 1 hr and 45 min long.
Click here to learn more
NOW YOU CAN BUILD, HOST, AND
SHARE APPS ON CLAUDE
Anthropic has introduced a major new feature for its
Claude AI platform: the ability to build, host, and
share interactive AI-powered apps directly within the
Claude app.
This capability, now available in beta to Free, Pro, and
Max users, dramatically lowers the barrier for creating
and distributing AI-driven tools, games, assistants,
and more
Click here to learn more
OWASP RELEASES AI TESTING GUIDE
This guide serves as the reference standard for
testing AI applications, going beyond traditional
software testing to address the distinct challenges of
AI, such as non-deterministic behavior, data-driven
risks, and adversarial threats.
Aimed at developers, security professionals,
architects, data analysts, researchers, and risk
officers, ensuring AI risks are addressed throughout
the product lifecycle
Click here to learn more
OXFORD UNIVERSITY’S SURVEY
ABOUT DEEP RESEARCH AGENTS
The paper systematically analyzes the technologies,
architectures, workflows, and challenges that define
the emerging field of Deep Research (DR) agents—
advanced autonomous AI systems designed to
perform complex, multi-turn research tasks using
large language models (LLMs) as their core
Click here to learn more
MEM0 RELEASES OPENMEMORY
CHROME EXTENSION
Mem0's OpenMemory Chrome Extension is a free,
client-side tool designed to provide a universal,
persistent memory layer for browser-based AI
assistants, such as ChatGPT, Claude, Perplexity,
Grok, Gemini, and others.
Its main goal is to solve the problem of context
fragmentation—where information shared with one
AI tool is not remembered by another.
Click here to learn more
POSTMAN'S "AI-READY APIS"
INITIATIVE
This new initiative is a suite of tools and capabilities
designed to help teams build, test, and govern APIs
that are optimized for integration with AI agents and
modern AI-driven systems.
Some of the new feature includes:
Generate MCP servers directly from Postman
collections, making APIs callable by AI agents.
Click here to learn more
SALESFORCE FOUNDER ENVISIONS A
FUTURE WITH 1 BILLION AI AGENTS
Benioff shares that Salesforce is on track to reach its
target of one billion AI agents, with 5,000 customers
already deploying these agents and rapid growth in
agent creation.
He also shares that AI now handles 30–50% of work in
key Salesforce functions like engineering, coding, and
support, and expects this trend to continue.
Click here to learn more
HEYGEN INTRODUCES HEYGEN
VIDEO AI AGENT
HeyGen Video Agent is a groundbreaking AI-
powered tool designed to automate the entire video
creation process from a single prompt or a set of
user-provided materials.
Marketed as the world’s first "prompt-native creative
engine," it transforms ideas, scripts, or uploaded
media into fully produced, publish-ready videos
Click here to learn more
GITHUB COPILOT CODING AGENT
IS AVAILABLE FOR PRO USERS
The GitHub Copilot coding agent is now available in
public preview for Copilot Pro users.
This new feature introduces an autonomous AI agent
capable of handling coding tasks directly from
GitHub issues, streamlining and automating parts of
the software development workflow
Click here to learn more
ELEVENLABS INTRODUCES [Link]
[Link] is a next-generation, voice-first AI assistant
developed by the team.
It has real-time, natural conversations with minimal
delay, making interactions feel immediate and fluid.
It even uses the MCP—to securely connect with
external apps and services like Slack, Google
Calendar, Perplexity, Salesforce, HubSpot, Gmail,
Zapier, and more
Click here to learn more
ANTHROPIC RELEASES EFFECT OF
CLAUDE ON COMPANIONSHIP
Anthropic presents the company’s first large-scale
analysis of how users turn to its Claude chatbot for
emotional support, interpersonal advice, and
companionship.
They share key findings like: Users seek help with
career development, relationship navigation,
loneliness, and existential questions about purpose
and meaning
Click here to learn more
DidImissany
Updates?
Let me know in the comments
below