Microsoft announces general availability of Azure DevOps local MCP Server
The MCP server acts as an intermediary between AI assistants and the Azure DevOps organization. It can help inject context from Azure DevOps, like work items, pull requests, and test plans. By adding this context to prompts, the LLM can provide more relevant answers that are tailored to the users Azure DevOps project.
According to Microsoft, it supports most of the main areas of Azure DevOps, including work items, wiki, repos, search, and test plans. Because it is local and not a remote MCP Server, it runs inside the network or local development environment, ensuring that private data doesn’t leave the system.
Google introduces Coral NPU, Veo 3.1 in Gemini API, and interactive commands in Gemini CLI
Google this week announced that the latest version of its image generation model Veo is now available in the Gemini API. Veo 3.1 can generate richer native audio, has an improved understanding of cinematic styles, and features enhanced image-to-video capabilities.
The company also announced Coral NPU, an open-source hardware architecture for edge AI that aims to address the core problems of performance, fragmentation, and privacy on low-power devices, like wearables. The architecture consists of a lightweight, C-programmable RISC-V frontend for managing data flow; a single instruction multiple data co-processor compatible with the RISC-V Vector instruction set 1.0; and a matrix execution unit.
Finally, the company added interactive commands in the Gemini CLI, including vim for editing, top for monitoring, and git rebase -i. With these additions, developers will have to do less switching between different terminal windows.
Anthropic releases Claude Haiku 4.5, a cost effective alternative to Claude Sonnet 4 and 4.5
Anthropic has announced the release of Claude Haiku 4.5, which provides similar coding performance to Claude Sonnet 4, but is twice as fast and one-third as expensive.
This follows the release of Claude Sonnet 4.5 two weeks ago, which Anthropic claimed as “the best coding model in the world” when it was announced. Claude Haiku 4.5 outperforms Sonnet 4 on SWE-bench, but falls second behind Sonnet 4.5.
“Claude Sonnet 4.5, released two weeks ago, remains our frontier model and the best coding model in the world. Claude Haiku 4.5 gives users a new option for when they want near-frontier performance with much greater cost-efficiency,” Anthropic wrote in a post.
BrowserStack adds Visual Review Agent for web testing
Software testing company BrowserStack announced a Visual Review Agent for its visual web testing platform Percy.
“Visual testing has reached a point where the question is no longer ‘why’ but ‘how to make it scalable?’” Ritesh Arora, CEO and co-founder at BrowserStack “The Visual Review Agent brings AI into the heart of visual testing, cutting through noise and surfacing what matters.”
According to BrowserStack, the Visual Review Agent provides an AI-generated summary of each visual change.
The agent uses AI to filter out minor pixel shifts and only flags meaningful changes, not minor changes like font resizes, margin shifts, or button alignment. “These noisy false positives force teams to waste time reviewing irrelevant changes instead of focusing on what actually matters,” the company explained in a demo
Salesforce announces general availability of its unified agentic AI platform Agentforce 360
Agentforce was first introduced last October, and over the last year, the company has added several new capabilities to the platform, such as the ability to embed agents in workflows and improved interoperability and governance.
“These milestones have led to Agentforce 360, which transforms Salesforce beyond CRM — where agents don’t just serve customers, but also empower employees, streamline operations, and collaborate with other agents,” Salesforce wrote in an announcement.
Agentforce 360 includes the Agentforce Builder, a development platform for building, testing, and deploying agents using natural language. It includes a unified workspace for all these phases of the development process, the ability to switch between a doc-like editor and low-code canvas, real-time previews and live debugging, and enterprise-grade governance.
Oracle launches AI Agent Marketplace to enable customers to find and deploy validated agents
Oracle has announced the launch of a new marketplace to help customers find and deploy validated AI agents that can be used within Oracle Fusion Applications, its suite of cloud-based modular business applications.
The AI Agent Marketplace includes agent templates built by Oracle partners to help with various business processes, including finance, HR, supply chain, and customer experience.
For example, IBM created a Smart Sales Order Entry Assistant agent to help cut down on manual order entry errors and accelerate order capture, Box offers a Data Extraction agent for the Oracle Digital Assistant that extracts metadata or structured content from documents stored in Box, and Stripe built an Infosys Invoice Collection AI that can help with payment processing and reminder.
