Google’s Agent2Agent protocol finds new dwelling on the Linux Basis
On the Open Supply Summit North America, it was introduced that Google donated its Agent2Agent (A2A) protocol to the Linux Basis.
The A2A protocol affords an ordinary approach for connecting brokers to one another. On this approach, it enhances Anthropic’s Mannequin Context Protocol (MCP), which offers a technique to join brokers to completely different knowledge sources and purposes.
“Drawing on Google’s inner experience in scaling agentic methods, we designed the A2A protocol to deal with the challenges we recognized in deploying large-scale, multi-agent methods for our clients. A2A empowers builders to construct brokers able to connecting with another agent constructed utilizing the protocol and affords customers the pliability to mix brokers from varied suppliers,” Google wrote in a weblog publish when it first launched A2A in April.
OpenAI provides Deep Analysis and Webhooks to the API
The addition of Deep Analysis will allow builders to construct analysis brokers that discover, analyze, and synthesize knowledge.
Webhooks have been additionally added, enabling builders to obtain notifications for API occasions like accomplished responses, fine-tuning jobs, and batch jobs.
Moreover, the corporate is dropping the worth for internet search and including it into extra fashions. It prices $10 / 1k instrument calls in o3, o3-pro, and o4 mini, and $25 / 1k instrument calls in GPT-4o and GPT-4.1.
Anthropic provides capacity to host and share Claude apps in its platform
Now, builders will be capable of not solely work together with Claude, but in addition use it to construct, host, and share their creations, eliminating the necessity to fear about internet hosting it themselves.
Customers will authenticate with their very own Claude account, and their API utilization will depend towards their subscription as an alternative of the app developer being charged.
Qodo launches CLI agent framework
Qodo, maker of an AI coding platform, in the present day introduced the discharge of Qodo Gen CLI, an agent framework that permits builders to create, customise, and deploy their very own AI coding brokers.
With the framework, creating brokers could be carried out by writing configuration information that add autonomous AI brokers all through the software program improvement life cycle, in keeping with the corporate’s announcement.
Qodo was constructed to assist builders add autonomous coding capabilities to their purposes with out requiring experience in AI methods, which might result in options that sync up with a company’s necessities, the corporate stated. With Qodo Gen CLI, builders can outline customized brokers and what instruments they’ll entry, specify actions that set off the brokers, what directions information their habits and finally, what their outputs needs to be.
Warp 2.0 evolves terminal expertise into an Agentic Growth Atmosphere
Warp is present process a big transformation with its 2.0 launch, shifting from its origins as a terminal emulator with AI integrations into an Agentic Growth Atmosphere (ADE).
It consists of 4 foremost capabilities: Code, Brokers, Terminal, and Drive. Any of these could be initiated from the primary interface, which accepts each prompts and terminal instructions.
“The merchandise available on the market in the present day, from AI IDEs to CLI coding brokers, all miss the mark supporting this workflow. They bolt brokers onto code editors by means of chat panels and bury them in CLI apps. What’s wanted is a product native to the agentic workflow; one primarily designed for prompting, multi-threading, agent administration, and human-agent collaboration throughout real-world codebases and infrastructure,” Zach Lloyd, the corporate’s CEO and founder, wrote in a weblog publish.
Agent Mode for Gemini added to Android Studio
With Agent Mode, a developer can describe a fancy aim, then the agent will give you an execution plan after which full the duties.
Examples of duties Agent Mode can deal with embrace constructing a challenge and fixing errors, extracting hardcoded strings and migrating them to strings.xml, including assist for darkish mode to an app, and implementing a brand new display in an app from a screenshot.
Builders could have the flexibility to evaluate, settle for, or reject any of the agent’s proposed modifications, or ask it to iterate on their suggestions. There may be additionally an auto-approve characteristic that may be enabled for conditions when a developer needs to iterate shortly on concepts.
Vercel Agent launches in restricted beta
The Vercel Agent is an AI assistant that analyzes Vercel app efficiency and safety knowledge.
It may possibly summarize anomalies, determine doubtless root causes, and suggest remediation actions throughout your complete platform, from managing firewall guidelines to figuring out optimization alternatives.
Tricentis Agentic Take a look at Automation
This can be a new AI agent that may generate take a look at instances robotically, leveraging text-based prompts in addition to prior take a look at runs. It additionally makes use of Tricentis’ Imaginative and prescient AI know-how to interpret visible parts throughout platforms, and integrates with Tricentis Tosca.
Moreover, the corporate launched a distant MCP server and a beta for its AI workflows functionality that permits higher communication between brokers and people.
Gemini 2.5 Professional and Flash are typically out there and Gemini 2.5 Flash-Lite in preview
In line with Google, no modifications have been made to Professional and Flash because the final preview, aside from the pricing for Flash is completely different. When these fashions have been first introduced, there was separate pondering and non-thinking pricing, however Google stated that separation led to confusion amongst builders.
The brand new pricing for two.5 Flash is identical for each pondering and non-thinking modes. The costs are actually $0.30/1 million enter tokens for textual content, picture, and video, $1.00/1 million enter tokens for audio, and $2.50/1 million output tokens for all. This represents a rise in enter price and a lower in output price.
Google additionally launched a preview of Gemini 2.5 Flash-Lite, which has the bottom latency and value among the many 2.5 fashions. The corporate sees this as an economical improve from 1.5 and a couple of.0 Flash, with higher efficiency throughout most evaluations, decrease time to first token, and better tokens per second decode.
Gemini 2.5 Flash-Lite additionally permits customers to regulate the pondering funds through an API parameter. Because the mannequin is designed for price and velocity effectivity, pondering is turned off by default.
GitHub Copilot Areas arrive
GitHub Copilot Areas permit builders to bundle the context Copilot ought to learn right into a reusable house, which might embrace issues like code, docs, transcripts, or pattern queries.
As soon as the house is created, each chat, completion, or command Copilot works from will probably be grounded in that information, enabling it to provide “solutions that really feel like they got here out of your group’s resident skilled as an alternative of a generic mannequin,” GitHub defined.
Copilot Areas will probably be free throughout its public preview and gained’t depend towards Copilot seat entitlements when the bottom mannequin is used.
OpenAI improves prompting in API
The corporate has now made it simpler to reuse, share, save, and handle prompts within the API by making prompts an API primitive.
Prompts could be reused throughout the Playground, API, Evals, and Saved Completions. The Immediate object will also be referenced within the Responses API and OpenAI’s SDKs.
Moreover, the Playground now has a button that can optimize the immediate to be used within the API.
“By unifying prompts throughout our surfaces, we hope these modifications will assist you to refine and reuse prompts higher—and extra promptly,” OpenAI wrote in a publish.
Syncfusion releases Code Studio
Code Studio is an AI-powered code editor that differs from different choices out there by having the LLM make the most of Syncfusion’s library of over 1,900 pre-tested UI parts fairly than producing code from scratch.
It affords 4 completely different help modes: Autocomplete, Chat, Edit, and Agent. It really works with fashions from OpenAI, Anthropic, Google, Mistral, and Cohere, in addition to self-hosted fashions. It additionally comes with governance capabilities like role-base entry, audit logging, and an admin console that gives utilization insights.
“Code Studio started as an in-house instrument and in the present day writes as much as a 3rd of our code,” stated Daniel Jebaraj, CEO of Syncfusion. “We created a safe, model-agnostic assistant so enterprises can plug it into their stack, faucet our confirmed UI parts, and ship cleaner options in much less time.”
AI Alliance splits into two new non-profits
The AI Alliance is a collaborative effort amongst over 180 organizations throughout analysis, educational, and trade, together with Carnegie Mellon College, Hugging Face, IBM, and Meta. It has now been included right into a 501(c)(3) analysis and training lab and a 501(c)(6) AI know-how and advocacy group.
The analysis and training lab will concentrate on “managing and supporting scientific and open-source initiatives that allow open neighborhood experimentation and studying, main to raised, extra succesful, and accessible open-source and open knowledge foundations for AI.”
The know-how and advocacy group will concentrate on “world engagement on open-source AI advocacy and coverage, driving know-how improvement, trade requirements and finest practices.”
Digital.ai introduces Fast Defend Agent
Fast Defend Agent is a cell software safety agent that follows the suggestions of OWASP MASVS, an trade normal for cell app safety. Examples of OWASP MASVS protections embrace obfuscation, anti-tampering, and anti-analysis.
“With Fast Defend Agent, we’re increasing software safety to a broader viewers, enabling organizations each giant and small so as to add highly effective protections in just some clicks,” stated Derek Holt, CEO of Digital.ai. “In in the present day’s AI world, all apps are in danger, and by democratizing our app hardening capabilities, we’re enabling the safety of extra purposes throughout a broader set of industries. With eighty-three % of purposes underneath fixed assault – the continued innovation inside our core choices, together with the launch of our new Fast Defend Agent, couldn’t be coming at a extra essential time.”
IBM launches new integration to assist unify AI safety and governance
It’s integrating its watsonx.governance and Guardium AI safety options in order that corporations can handle each from a single instrument. The built-in answer will be capable of validate towards 12 completely different compliance frameworks, together with the EU AI Act and ISO 42001.
Guardium AI Safety is being up to date to have the ability to detect new AI use instances in cloud environments, code repositories, and embedded methods. Then, it may possibly robotically set off the suitable governance workflows from watsonx.governance.
“AI brokers are set to revolutionize enterprise productiveness, however the very advantages of AI brokers may current a problem,” stated Ritika Gunnar, common supervisor of knowledge and AI at IBM. “When these autonomous methods aren’t correctly ruled or secured, they’ll carry steep penalties.”
Safe Code Warrior introduces AI Safety Guidelines
This new ruleset will present builders with steering for utilizing AI coding assistants securely. It permits them to determine guardrails that discourage the AI from dangerous patterns, resembling unsafe eval utilization, insecure authentication flows, or failure to make use of parameterized queries.
They are often tailored to make use of with quite a lot of coding assistants, together with GitHub Copilot, Cline, Roo, Cursor, Aider, and Windsurf.
The foundations can be utilized as-is or tailored to an organization’s tech stack or workflow in order that AI-generated output higher aligns throughout initiatives and contributors.
“These guardrails add a significant layer of protection, particularly when builders are transferring quick, multitasking, or discover themselves trusting AI instruments a bit of an excessive amount of,” stated Pieter Danhieux, co-founder and CEO of Safe Code Warrior. “We’ve stored our guidelines clear, concise and strictly centered on safety practices that work throughout a variety of environments, deliberately avoiding language or framework-specific steering. Our imaginative and prescient is a future the place safety is seamlessly built-in into the developer workflow, no matter how code is written. That is only the start.”
SingleStore provides new capabilities for deploying AI
The corporate has improved the general knowledge integration expertise by permitting clients to make use of SingleStore Circulate inside Helios to maneuver knowledge from Snowflake, Postgres, SQL Server, Oracle, and MySQL to SingleStore.
It additionally improved the mixing with Apache Iceberg by including a velocity layer on high of Iceberg to enhance knowledge alternate speeds.
Different new options embrace the flexibility for Aura Container Service to host Cloud Features and Inference APIs, integration with GitHub, Notebooks scheduling and versioning, an up to date billing forecasting UI, and simpler pipeline monitoring and sequences.
Apple pronounces Basis Fashions framework
The Basis Fashions framework will allow any app developer to make the most of Apple Intelligence’s on-device processing for his or her apps. Apple Intelligence leverages Apple’s silicon processors to have the ability to course of knowledge with out it leaving the gadget, making certain higher knowledge privateness for customers.
In line with Apple, Automattic’s journaling app Day One utilized the framework so as to add intelligence options which are privacy-centric.
“The Basis Mannequin framework has helped us rethink what’s potential with journaling,” stated Paul Mayne, head of Day One at Automattic. “Now we will carry intelligence and privateness collectively in ways in which deeply respect our customers.”
The framework natively helps Swift, and contains capabilities resembling guided era and gear calling.
Mistral launches its first reasoning mannequin
Magistral is a reasoning mannequin that excels in “domain-specific, clear, and multilingual reasoning,” in keeping with the corporate.
It is available in two completely different sizes: Magistral Small is a 24B parameter model and Magistral Medium is a extra highly effective model for the enterprise. Magistral Small can be open supply in order that the neighborhood can construct upon its structure and reasoning processes.
New Relic provides MCP assist in AI Monitoring answer
What this implies is that builders who’re constructing brokers with MCP will be capable of get insights into the life cycle of an MCP request, which encompasses invoked instruments, name sequences, and execution durations.
With this new functionality, builders will be capable of correlate MCP efficiency with your complete software ecosystem. It can allow them to trace utilization patterns, latency, errors, and efficiency, to allow them to optimize their MCP providers.
ArmorCode releases AI Code Insights
This newest providing makes use of the corporate’s agent Anya, which has contextual understanding of a buyer’s code repositories.
Its key advantages embrace the flexibility to raised prioritize remediation efforts, floor hidden belongings, handle change dangers, perceive code possession, and have proactive AI publicity administration.
“Whereas improvement velocity has skyrocketed, safety groups are sometimes flying blind, buried in alerts with out understanding the precise threat lurking inside their code repositories,” stated Mark Lambert, chief product officer at ArmorCode. “AI Code Insights modifications that. We’re offering the essential context – the ‘what, who and the way’ – behind the code and vulnerability. This permits organizations to lastly lower by means of the noise, prioritize successfully, and proactively safe their most crucial belongings earlier than they grow to be liabilities. It’s about making present safety investments work smarter, not simply more durable.”
Amplitude launches AI brokers for product improvement
These new brokers can help with bettering checkout conversion, characteristic adoption, and consumer onboarding. It may possibly additionally search for indicators {that a} consumer is able to improve in order that it may possibly ship them particular affords.
With Amplitude’s AI Brokers working across the clock, product improvement shifts from a gradual, step-by-step course of to a high-speed, multi-track system the place technique, evaluation, and motion can occur on the identical time,” stated Spenser Skates, CEO and co-founder of Amplitude. “This isn’t nearly doing what you’ve at all times carried out, quicker. It’s about doing what you wouldn’t, couldn’t, or didn’t know learn how to do earlier than.”
OpenAI o3-pro out there within the API
OpenAI has introduced that o3-pro can now be accessed by means of the OpenAI API. Moreover, Professional and Crew customers can now use o3-pro in ChatGPT.
The pricing for utilizing o3-pro within the API will probably be 87% cheaper than o1-pro, and the worth of o3 can be being lower by 80%. “We optimized our inference stack that serves o3. Identical precise mannequin—simply cheaper,” the corporate wrote in a publish on X.
Zencoder launches end-to-end UI testing agent
Zencoder has introduced a public beta for Zentester, its new end-to-end UI testing AI agent.
Zentester imitates how people behave when interacting with internet purposes, resembling navigating the format, and figuring out and utilizing interactive parts. It does this by combining photographs (screenshots) with DOM (snapshot) data.
Because it runs by means of take a look at eventualities, it generates take a look at artifacts that seize the actions carried out and the anticipated visible and practical outcomes.
Databricks provides new instruments like Lakebase, Lakeflow Designer, and Agent Bricks to raised assist constructing AI apps and brokers within the enterprise
Lakebase is a managed Postgres database designed for operating AI apps and brokers. It provides an operational database layer to Databricks’ Information Intelligence Platform.
Coming quickly as a preview, Lakeflow Designer is a no-code ETL functionality for creating manufacturing knowledge pipelines. It contains a drag-and-drop UI and an AI assistant that enables customers to explain what they need in pure language.
It’s based mostly on Lakeflow, the corporate’s answer for knowledge engineers for constructing knowledge pipelines. Lakeflow is now typically out there, with new options resembling Declarative Pipelines, a brand new IDE, new point-and-click ingestion connectors for Lakeflow Join, and the flexibility to jot down on to the lakehouse utilizing Zerobus.
Agent Bricks is Databricks’ new instrument for creating brokers for enterprise use instances. Customers can describe the duty they need the agent to do, join their enterprise knowledge, and Agent Bricks handles the creation.
Trustwise launches belief layer for AI brokers
Its new Concord AI answer contains six shields that safe AI throughout completely different fashions, brokers, and clouds. They embrace an MCP Protect, Immediate Protect (to forestall injection assaults and hallucinations), Compliance Protect, Model Protect (for sustaining model tone and persona), Price Protect, and Carbon Protect (for decreasing carbon footprint of AI).
“Builders aren’t simply securing textual content anymore, they’re securing actions,” stated Matthew Barker, head of AI analysis at Trustwise. “That calls for real-time controls that assist each builders and safety groups monitor how brokers assume, resolve, and act. Concord AI acts as a runtime defend, implementing safety and management straight within the resolution loop and stopping drift earlier than brokers go off target.”
OpenAI pronounces Codex updates
The coding agent Codex can now entry the web throughout activity execution, opening up new capabilities resembling the flexibility to put in base dependencies, run exams that want exterior assets, and improve or set up packages.
Web entry is turned off by default. It may be enabled when a brand new setting is created, or an setting could be edited to permit it. Customers can management the domains and HTTP strategies that Codex can use.
OpenAI additionally introduced that Codex has begun rolling out to ChatGPT Plus customers. The corporate did word that it would set price limits for Plus customers throughout excessive demand intervals.
Mistral releases coding assistant
Mistral Code builds on the open-source challenge Proceed, which offers a hub of fashions, guidelines, prompts, docs, and different constructing blocks for creating AI code assistants. It’s powered by 4 completely different coding fashions: Codestral, Codestral Embed, Devstral, and Mistral Medium.
It’s proficient in over 80 programming languages, and might cause over information, Git diffs, terminal output, and points. It’s presently out there as a personal beta for JetBrains IDEs and VSCode.
“Our aim with Mistral Code is straightforward: ship best-in-class coding fashions to enterprise builders, enabling all the pieces from instantaneous completions to multi-step refactoring—by means of an built-in platform deployable within the cloud, on reserved capability, or air-gapped on-prem GPUs. Not like typical SaaS copilots, all elements of the stack—from fashions to code—are delivered by one supplier topic to a single set of SLAs, and each line of code resides contained in the buyer’s enterprise boundary,” the corporate wrote in its announcement.
Postman introduces Agent Mode to combine the facility of AI brokers into Postman’s core capabilities
The brokers can create, set up, and replace collections; create take a look at instances; generate documentation; construct multi-step brokers to automate repeatable API duties; and setup monitoring and observability.
Abhinav Asthana, CEO and co-founder of Postman, informed SD Occasions that it’s type of like having an skilled Postman consumer by your facet.
The corporate additionally introduced the flexibility for customers to show any public API on the Postman community into an MCP server. It additionally launched a community for MCP servers the place publishers can host instruments for brokers and have them be simply discoverable by builders. “We principally took all of the distant MCP servers out there in the present day, verified them, and put them on the general public community,” stated Abhinav Asthana, CEO and co-founder of Postman.
FinOps Basis launches FinOps for AI certification
The coaching and certification is designed to “assist FinOps practitioners perceive, handle, and optimize AI-related cloud spend,” the inspiration defined.
It can tackle subjects resembling AI-specific price allocation, chargeback fashions, workload optimization, unit economics, and sustainability.
The teachings will probably be a four-part collection, the primary of which is now out there, with the opposite elements launching in September 2025, November 2025, and January 2026. The certification examination will probably be out there in March of subsequent 12 months.
Newest model of Microsoft’s Dev Proxy provides LLM utilization and value monitoring
Dev Proxy 0.28 contains the OpenAITelemetryPlugin to supply visibility into how purposes are interacting with OpenAI. For every request, it’ll present details about the mannequin used, the token depend, price estimation, and grouped summaries per mannequin.
Dev Proxy may now use the native AI runtime stack Foundry Native as its native language mannequin supplier.
Different updates in Dev Proxy 0.28 embrace new extensions for .NET Aspire, improved producing PATCH operations for TypeSpec, assist for JSONC in mock information, and improved logging.
Snowflake introduces agentic AI improvements for knowledge insights
Snowflake Intelligence (public preview quickly) is powered by clever knowledge brokers and offers a pure language expertise for asking questions that can lead to the supply of actionable insights from structured and unstructured knowledge. Additionally in personal preview quickly is a brand new Information Science Agent to assist knowledge scientists automate routine ML mannequin improvement duties, in keeping with the corporate’s announcement.
Snowflake Intelligence brings collectively knowledge from quite a few sources and makes use of the brand new Snowflake Openflow to compile data from spreadsheets, paperwork, photographs, and databases concurrently. The info brokers can generate visualizations and help customers in taking motion on insights, Snowflake stated in its announcement. Snowflake Intelligence may entry third-party information by means of Cortex Information Extensions, quickly to be typically out there on Snowflake Market.
Progress provides new AI code assistants
Progress Software program introduced new AI code assistants and different capabilities constructed into the Q2 2025 launch of Progress Telerik and Progress Kendo UI, .NET and JavaScript UI libraries for contemporary software improvement. This launch introduces AI Coding Assistants for Blazor and React, AI-driven theme era and GenAI-powered reporting insights, the corporate introduced.
The AI Coding Assistants can robotically generate code in Telerik UI for Blazor and KendoReact libraries inside many AI-powered IDEs, which reduces the time spent on guide edits and shortens improvement cycles. Additional, developerws could make pure language prompts into Progress ThemeBuilder to create customized types for Telerik and Kendo UI parts, the corporate wrote in its announcement. The releases additionally embrace reporting summaries and insights powered by generative AI in Progress Telerik Reporting, in addition to a GenAI-powered PDF processing library, which the corporate stated can present “instantaneous doc insights, AI immediate choices within the Editor management and new AI constructing blocks and web page templates to hurry up UI improvement.”
IBM pronounces wastonx AI Labs
Watsonx AI Lab is an innovation hub based mostly in New York Metropolis that can join AI builders with IBM’s assets and experience.
In line with IBM, NYC was chosen for the situation as a result of it has over 2,000 AI startups. The corporate hopes to assist these in addition to pursue collaborations with native universities and analysis establishments.
“This isn’t your typical company lab. watsonx AI Labs is the place the most effective AI builders achieve entry to world-class engineers and assets and construct new companies and purposes that can reshape AI for the enterprise,” stated Ritika Gunnar, common supervisor of knowledge and AI at IBM. “By anchoring this mission in New York Metropolis, we’re investing in a various, world‑class expertise pool and a vibrant neighborhood whose improvements have lengthy formed the tech panorama.”
