Latest news across AI, security, and the OpenClaw ecosystem.
Agmsg is a new bash and SQLite tool that lets CLI AI agents such as Claude Code, Codex, and Gemini CLI send messages to one another through a shared local database. The project uses no daemon or network and supports real-time or between-turns delivery depending on the agent.
Source: HN Show HNA new free plugin adds local audio and spoken alerts for Claude Code and OpenAI Codex. It runs offline, uses built-in operating system voices and sounds, and can notify users when a task finishes or needs a decision.
Source: HN Show HNHermes Agent v0.15.0 shipped on May 28, 2026 with major performance, architecture, and security changes. The release cuts run_agent.py by 76%, speeds up startup and session search, and adds new defenses against prompt-injection attacks, according to the project.
Source: Hermes Agent ReleasesOpenAI tagged rust-v0.135.0 of Codex on GitHub at commit 4daceea. The release page does not include a changelog in the provided source, so the update is mainly a version and commit reference for the Codex Rust branch.
Source: Codex ReleasesData Formulator 0.7 is an open-source AI-powered data analytics system for enterprise teams. The release adds reusable Data Connectors, context-aware agents, and an interactive workspace for exploring and refining analyses across fragmented data sources.
Source: Microsoft ResearchAnthropic has added two security features to Claude: a self-hosted sandbox for Managed Agents and a security-guidance plugin that reviews code for common flaws during work. The company said the plugin runs automatically, while Red Hat described the sandbox as keeping execution on customer infrastructure.
Source: The Hacker NewsOpenClaw 2026.5.27 adds stronger security boundaries, blocks unsafe command and runtime settings, and improves delivery behavior across channels such as Slack, Telegram, iMessage, and Discord. The release also expands provider support, including OpenAI-compatible embeddings and Pixverse video generation, while hardening packaging, CI, and release verification.
Source: OpenClaw ReleasesEndava says it has adopted Codex across requirements, design, development, and operations, calling itself an “agentic organization.” The company says the tool helps junior staff work with senior-level guidance and has shortened some workflow steps from weeks to days.
Source: OpenAI NewsMicrosoft has introduced Windows 365 for Agents, a cloud PC platform in public preview for running AI agents in secure enterprise environments. The company says organizations can use existing controls in Entra ID and Intune to govern agent activity across apps, browsers, files, and legacy systems.
Source: Help Net SecurityAnthropic has introduced dynamic workflows in Claude Code, a research preview feature that breaks large tasks into parallel subagent runs. The company says it can help with codebase-wide bug hunts, migrations, and other long-running engineering work across Claude Code, the API, and major cloud platforms.
Source: Claude BlogAnthropic has released Claude Opus 4.8, an updated model that the company says improves on Opus 4.7 across coding, reasoning, and agent tasks. The launch also adds dynamic workflows in Claude Code, effort controls in claude.ai, and lower fast-mode pricing for Opus 4.8.
Source: Anthropic NewsVAEN is a CLI that packages agent.yaml-based coding-agent setups into OCI-backed .agent archives and imports them into other repositories without secrets. It validates manifests, inspects bundle contents, and writes client-specific files for Codex, Claude Code, or Copilot.
Source: HN Show HNKilo Code users can now sign in with an X Premium+ or SuperGrok subscription to use Grok models inside the platform. According to the company, the integration works without a separate API key and includes support for Grok Build 0.1 for agentic coding workflows.
Source: Kilo BlogGoogle Cloud introduced AI Threat Defense, an automated security platform that finds, prioritizes, and helps patch software flaws using its Gemini models, Wiz, CodeMender, and Mandiant. The company says the system is built for a threat environment where attackers use AI to find and exploit weaknesses in hours or days.
Source: Help Net SecurityAnthropic has added a security-guidance plugin to Claude Code that reviews code for common vulnerabilities while developers work. The company says it can catch issues before pull requests, and that internal use reduced security-related PR comments by 30% to 40%.
Source: Help Net SecurityOpenClaw’s 2026.5.26 update adds a faster Gateway path, broader transcript handling, and stronger content-boundary checks. The release also improves Telegram, iMessage, WhatsApp, Discord, Signal, voice features, installs, and diagnostics.
Source: OpenClaw ReleasesCisco says OpenAI’s Codex is being used across its engineering workflows, including AI Defense, build optimization, defect repair, and framework migrations. The company said the tool helped cut some work from weeks or quarters to days or hours.
Source: OpenAI NewsOpenAI says it co-developed Tax AI with Thrive Holdings and Crete accountants to automate tax preparation and improve the system through production feedback. The company says the tool processed 7,000 returns in a pilot and improved from launch to the point where more returns reached higher accuracy thresholds over six weeks.
Source: OpenAI NewsMicrosoft is adding security and governance features for AI agents as enterprises deploy more AI tools and connectors. PwC says the goal is to help companies move quickly without accumulating new security risks, or “AI security debt.”
Source: CIO AIWarp says it is using GPT-5.5 to support open-source agentic development workflows, where agents write code and humans supervise outcomes. The company says GPT-5.5 uses 30% fewer tokens than GPT-5.4 in internal benchmarks and helps power its Oz orchestration platform.
Source: OpenAI NewsA new macOS app called Vision Clicker watches a selected screen region and automatically clicks approval buttons such as Run, Fetch, or Retry using Apple Vision OCR. The tool runs locally, needs Accessibility and Screen Recording permissions, and is intended for small approval prompts in coding-agent UIs.
Source: HN Show HNHarbor v0.4.19 adds a new `harbor launch` command for starting pre-configured local coding tools and pointing them at inference backends. The release also adds web search and URL-reading support through Harbor Boost, plus compatibility and hardening updates.
Source: HN Show HN7AI has launched PLAID ELITE, a fully managed AI-native security operations service that uses autonomous agents with oversight from 7AI engineers. The company says it can handle alert triage, investigation and response continuously, while reducing false positives and speeding investigations from hours to minutes.
Source: SiliconANGLEAnthropic said Claude Managed Agents can now run in self-hosted sandboxes and connect to private Model Context Protocol, or MCP, servers through MCP tunnels. The company announced the features at Code w/ Claude London 2026, and said teams including Amplitude, Clay and Rogo are already using the managed agents setup.
Source: Claude BlogOpenClaw passed 300,000 GitHub stars in April, while Google launched Gemini Spark at I/O. The two products do similar tasks, but OpenClaw runs on user-owned hardware and Spark runs on Google Cloud.
Source: The New StackAnthropic says its Project Glasswing program has helped uncover more than 10,000 high- or critical-severity vulnerabilities since launching last month. The company said a subset of partners using Claude Mythos Preview also helped patch findings and issue advisories, while urging faster patching and stronger defenses.
Source: The Hacker NewsLilo is an open source personal AI assistant that runs in Telegram and other channels, including email, WhatsApp, web, and mobile apps. Its creator says it is model-agnostic, keeps memory in a persistent workspace, and can use tools such as web search and browser control.
Source: HN Show HNMicrosoft Security highlighted customer stories from St. Luke’s University Health Network and ManpowerGroup, saying both organizations are building security foundations for AI. The examples focus on unified visibility, governance, and automation across cloud, identity, data, and operations.
Source: Microsoft Security BlogDell said open-source AI agents are changing how enterprises deploy and govern infrastructure, while the OpenClaw community is drawing new builders into software development. Dell executives said demand now spans from data center systems to deskside AI machines.
Source: SiliconANGLEAnthropic’s Claude for Small Business connected to Google Drive, Canva and Gmail to analyze a fictional seven-month P&L. In a test built with 20 planted problems, Claude found 17 in under six minutes and then created a Canva deck and email summary.
Source: The New StackVirgin Atlantic said it used OpenAI’s Codex to launch a revamped mobile app for the Christmas travel rush with near-complete unit test coverage and zero P1 defects. The airline also said Codex is speeding up legacy code refactors and helping analyst teams build internal tools on top of its data warehouse.
Source: OpenAI NewsCodex 0.133.0, released May 21, makes Goals enabled by default, adds a dedicated goal store, and changes remote-control behavior to run like a foreground command. The release also expands permission profiles, plugin discovery, and extension hooks, while fixing several app-server, TUI, and runtime issues.
Source: Codex ReleasesMicrosoft announced updates to Purview, Entra ID, and Windows 365 for Agents aimed at improving visibility and control across AI tools, data, and identities. The changes include a Claude connector for Purview, a new DSPM experience, OCR and custom examinations in investigations, and a secure execution environment for agents.
Source: Microsoft Security BlogWorkday has launched two AI agents for its Sana platform: one for IT service management and one for travel. The company says the tools will help automate onboarding, access provisioning, travel booking and expense reporting across HR, IT and finance.
Source: SiliconANGLEAnthropic said Claude now connects with 28 security and compliance tools through its Compliance API. The integrations give IT and security teams access to conversation content and activity events so they can apply existing monitoring, DLP, and governance policies to Claude.
Source: Claude BlogGoogle previewed Gemini Spark, a hosted personal AI agent that connects to Gmail, Calendar, Drive, Docs, Sheets, Slides, YouTube, and Google Maps. The company also said the open source Gemini CLI will stop working with Google AI subscription plans on June 18 and be replaced by the closed source Antigravity CLI.
Source: Simon WillisonMicrosoft has open-sourced two tools for AI agent development: RAMPART, a testing framework for continuous safety checks, and Clarity, a structured app for reviewing design assumptions before coding starts. The company says both tools are meant to make AI safety part of the normal engineering workflow.
Source: Microsoft Security BlogAtrium is a resumable workspace manager for Claude Code, Codex, and Gemini CLI that keeps agent sessions, panes, and task context alive across crashes and reboots. It uses isolated git worktrees for parallel work and offers a CLI for launching, steering, and reviewing agents.
Source: HN Show HNOpenClaw published release v2026.5.19 on GitHub. The source page shown for the release does not include changelog details or feature notes in the provided text.
Source: OpenClaw Releases1Password has released an MCP server for OpenAI’s Codex coding agent that lets secrets be accessed at runtime without exposing them in prompts, code or model context. The company said the integration uses just-in-time credentials and is part of its broader push to secure agentic development.
Source: SiliconANGLEGoogle is expanding Antigravity from a coding environment into a platform for managing autonomous AI agents. The update includes a desktop app, CLI, SDK, cloud integration, and new security tools such as CodeMender and an AI Content Detection API.
Source: The New StackGoogle unveiled Pics at I/O 2026, a Workspace-native AI image generator and editor powered by Nano Banana 2. The tool lets users generate images, then move, resize, translate, or edit individual elements without rebuilding the whole image.
Source: The Next WebElementalSoul has released claude-bughunter, a self-contained skill bundle for Claude Code aimed at bug hunting and external red-team work. The package includes 51 skills, 15 slash commands, and more than 574 disclosed-report patterns across 24 vulnerability classes.
Source: HN Show HNA new Chrome extension and local bridge let zot, a terminal-based coding agent, inspect and control browser tabs through a side panel. The project installs as a zot extension, then adds a zot-chrome command shim and connects to Chrome over a local WebSocket.
Source: HN Show HNCodex 0.132.0 adds first-class authentication to the Python SDK, including API key login and ChatGPT browser and device-code flows. The release also simplifies Python turn APIs, adds structured output support for resumed runs, and includes fixes for sessions, remote execution, Windows installs, and the terminal UI.
Source: Codex ReleasesAnthropic’s Travis Bryant says he uses Claude Cowork to prepare customer briefs, build weekly forecasts, and run an overnight scoring process for a 4,000-account territory. He says the system now handles data assembly and formatting work that previously took cross-functional teams hundreds of hours.
Source: Claude BlogRamp says engineers are using Codex with GPT-5.5 to get substantive pull request feedback in minutes and to build internal agentic tools faster. The company says the system helps reduce manual review work and supports its on-call automation project.
Source: OpenAI NewsGoogle announced new information agents for Search at its 2026 I/O keynote. The agents can monitor topics continuously, summarize updates from multiple sources, and send notifications to users, starting this summer for Google AI Pro and Ultra subscribers in the U.S.
Source: TechCrunchGoogle used I/O 2026 to announce two new models, Gemini Omni and Gemini 3.5 Flash, and to expand agent-style features across its products. The company said the updates move Gemini from a writing aid toward systems that can act on users’ behalf.
Source: Google AI BlogGoogle announced new AI features for Gmail, Docs, Keep, Drive and the Gemini app. The updates include voice-based tools, a new image editor called Google Pics, expanded AI Inbox features, and Gemini Spark, a personal AI agent that can act on a user’s behalf.
Source: Google AI Blog