AI Agents News
The latest news on AI agents and agentic AI — platform launches, framework releases, MCP updates, research, and industry moves. A focused view of our newsroom: every card links to the full story. 66 stories and counting.
OpenAI Adds Codex to ChatGPT Mobile App
OpenAI now allows users to control its Codex desktop AI tool directly from the ChatGPT app on iOS and Android devices. This preview feature enables phone-based management of tasks on the computer, including reviewing outputs and approving commands. It rolls out to all ChatGPT plans, amid OpenAI's efforts to compete with rivals like Anthropic's Claude Code.
Microsoft Deploys 100+ AI Agents to Find Windows Vulnerabilities
Microsoft launched MDASH, a security system with more than 100 specialized AI agents that detected 16 new Windows vulnerabilities, including four critical ones. The tool uses a four-stage process to analyze code and debate findings, achieving 88.45 percent on the CyberGym benchmark. Backed by experts from a DARPA challenge winner, MDASH is in limited preview.
Notion Launches Developer Platform for AI Agents
Notion has introduced a developer platform that turns its workspace into a center for AI agents. The platform includes Workers for custom code, database syncing from sources like Salesforce, and connections to external agents such as Claude Code. Teams can now build multi-step workflows and automate tasks without third-party tools.
Android Adds Gemini AI Agents for Trips, Forms, Texts
Google has launched new AI features under Gemini Intelligence for Android devices. These tools handle multi-step tasks like booking trips, filling forms, summarizing web pages, and refining spoken notes into text messages. The capabilities will first appear on Samsung Galaxy S26 and Google Pixel 10 this summer, expanding to other gadgets later.
Google Tests Remy AI Agent for Gemini App
Google has begun testing Remy, a new personal AI agent within the Gemini app, aimed at handling user tasks in work and daily life. The agent acts on behalf of users across Google services, with a focus on learning preferences and maintaining controls. Details come from an internal document and sources familiar with the project, though public release plans remain unclear.
Anthropic Releases 10 Finance Agent Templates
Anthropic has launched ten agent templates for key financial tasks like pitchbooks, KYC screening, and month-end closes. These work as plugins in Claude Cowork and Claude Code or cookbooks for Claude Managed Agents. Integration with Microsoft apps and new data connectors enhance capabilities, paired with Claude Opus 4.7 topping benchmarks.
Anthropic Launches 10 AI Agents for Finance Sector
Anthropic released ten preconfigured AI agents tailored for finance to handle tasks in research, risk checks, and accounting. These tools integrate with platforms like Claude Cowork and connect to new data partners including Moody's. The move comes as Anthropic and OpenAI seek enterprise revenue ahead of possible IPOs this year.
CopilotKit Raises $27M for App-Native AI Agents
Seattle startup CopilotKit has secured $27 million in Series A funding to advance tools for deploying AI agents directly in applications. The round, led by Glilot Capital, NFX, and SignalFire, supports development of its AG-UI protocol and enterprise toolkit. Backed by major providers like Google and Microsoft, the open-source protocol enables dynamic interfaces beyond text chats.
Sierra Raises $950M at Over $15B Valuation
Bret Taylor's AI startup Sierra announced a $950 million funding round led by Tiger Global and GV, resulting in a post-money valuation above $15 billion. The company now has more than $1 billion in capital to pursue its goal of setting the global standard for AI-powered customer experiences. Sierra reports rapid growth, serving over 40% of the Fortune 50, with agents handling billions of interactions, and ARR jumping from $100 million to $150 million in months.
OpenAI Releases Symphony for Autonomous AI Agents
OpenAI launched Symphony, an open-source specification that enables AI agents to manage tasks in tools like Linear without constant human oversight. Agents handle open tickets independently, create follow-ups, and allow developers to focus on reviews. Internal use showed merged pull requests increase sixfold in three weeks, with community adaptations already emerging.
DeepClaude Enables Claude Code Loop with DeepSeek V4 Pro, 17x Cheaper
DeepClaude lets users run Claude Code's autonomous agent loop using DeepSeek V4 Pro or other backends, cutting costs by 17 times compared to the original $200 monthly plan. It keeps the same terminal interface and features like file editing and bash execution while swapping the model backend. Setup takes two minutes with an API key, and it supports OpenRouter, Fireworks AI, and Anthropic options.
Mistral AI Unveils Medium 3.5 and Vibe Remote Agents
Mistral AI has released Mistral Medium 3.5, a 128B dense model for coding and productivity tasks, now the default in Vibe and Le Chat. New remote agents in Vibe allow cloud-based coding sessions started from CLI or Le Chat, running in parallel. Le Chat introduces Work mode for handling complex multi-step tasks with tool integration.
n8n Agents Join Teams as Members in Microsoft 365 Apps
n8n users can now create AI agents using Microsoft Agent 365 that appear as team members in Microsoft 365 applications such as Teams, Outlook, and Word. These agents receive their own Entra ID for access to services like SharePoint and Teams channels, managed through Microsoft tools for identity and compliance. The integration allows agents to handle tasks directly in conversations, pulling data from tools like Zendesk and Salesforce.
Stripe Launches Link Wallet for AI Agents
Stripe has unveiled Link, a digital wallet designed for an era of autonomous AI agents that handle shopping, reservations, and ticket purchases. Users can link payment methods, monitor spending, and manage subscriptions while securely authorizing AI agents to make payments. The wallet supports web, iOS, and Android, with features like OAuth integration for agent approvals and future expansions for limits and autonomous spending.
OpenAI Tells Codex to Stop Mentioning Goblins
OpenAI's instructions for its Codex coding agent explicitly prohibit discussions of goblins, gremlins, raccoons, trolls, ogres, pigeons, and other creatures unless directly relevant to user queries. This rule appears multiple times in the Codex CLI tool, amid reports of the AI fixating on such topics when powering OpenClaw, an acquired agentic tool. Users shared funny experiences on X, sparking memes, while OpenAI staff confirmed the issue ties to model behaviors in agent setups.
OpenAI Powers AWS Bedrock Managed Agents: CEOs Interview
OpenAI CEO Sam Altman and AWS CEO Matt Garman discussed Bedrock Managed Agents, powered by OpenAI models, in a recent interview. This follows Microsoft's amended agreement with OpenAI, allowing access on other clouds like AWS. The product packages OpenAI's frontier models in an AWS-native environment for enterprise agents handling identity, permissions, and more.
NVIDIA Launches Nemotron 3 Nano Omni for 9x Efficient AI Agents
NVIDIA has released Nemotron 3 Nano Omni, an open multimodal model that combines vision, audio, and language processing. It offers up to 9x higher throughput than other open omni models, topping leaderboards in document intelligence, video, and audio tasks. The model supports agentic workflows like computer use and is available now on multiple platforms.
Red Hat Engineer Releases Tank OS for Safer OpenClaw Use
Sally O'Malley, a Red Hat principal software engineer and OpenClaw maintainer, launched Tank OS on Tuesday. The open source tool simplifies safe deployment and management of OpenClaw agents for power users and IT teams. It uses Podman containers on Fedora Linux to isolate instances and handle enterprise-scale operations.
OpenAI Folds Codex into GPT-5.5, Ends Dedicated Model
OpenAI has integrated its separate Codex programming model into GPT-5.5, eliminating the standalone coding line that began with GPT-5.4. Romain Huet, Head of Developer Experience at OpenAI, notes that GPT-5.3 from early February marks the final independent Codex release. The new model shows advances in agentic coding, reduced token use, and higher API costs.
500 Bankers Review AI Outputs, None Fit for Clients
Around 500 investment bankers evaluated outputs from top AI models on junior banking tasks. None proved ready for client delivery. While 41 percent needed major changes and 27 percent were unusable, over half saw value as a starting point for work.
AI Agents Expand Software Engineering Beyond Code, Researchers Say
Researchers from Chalmers University of Technology and the Volvo Group claim AI agents will not replace software engineers. They introduce semi-executable artifacts like prompts and workflows that broaden the field. A new model called the semi-executable stack outlines six layers from code to societal rules.
Anthropic Tests Marketplace for AI Agent Commerce
Anthropic ran Project Deal, a pilot where AI agents acted as buyers and sellers for 69 employees with $100 budgets each. The experiment saw 186 deals worth over $4,000. Advanced models delivered better results, though participants did not notice the differences.
OpenAI Releases GPT-5.5 and GPT-5.5 Pro to API
OpenAI added GPT-5.5 and GPT-5.5 Pro to its Chat Completions and Responses APIs on April 23, 2026. These models support a 1M token context window, image inputs, and various tools. The changelog details numerous updates from 2023 to 2026, including image, video, audio models, agent tools, and API improvements.
OpenAI Launches Workspace Agents for Team Tasks
OpenAI provides workspace agents to Business, Enterprise, Edu, and Teachers plan users. These cloud-based tools in ChatGPT let teams create custom bots for independent work like reporting product feedback via Slack or drafting sales emails in Gmail. The agents mark an evolution from GPTs amid rising AI agent interest and competition from Anthropic.