Cursor adds event subscriptions, /goal, isolated subagent VMs, and non-interrupt steering
Cursor’s Aug 19 cloud-agent update adds event subscriptions, /goal, isolated subagent VMs, custom modes, and steering that waits for the next tool call.
News
Mastra documents first-class Vercel AI SDK v7 helpers and message conversion
Mastra documents AI SDK v7 helpers—chatRoute, workflowRoute, handle*Stream, and toAISdkStream—for TypeScript agent UIs.
Claude Agent SDK for Python 0.2.140 adds MCP 2.x and subagent transcript forwarding
Anthropic’s Python Agent SDK 0.2.140 adds MCP 2.x in-process servers and forward_subagent_text so nested transcripts can stream to the parent.
OpenAI packages Asana’s Enzyme-to-RTL Codex migration as a $12K, two-week case study
OpenAI’s Aug 18 Asana case study retells the Aug 7 Enzyme migration: four Codex agents, a five-sentence prompt, about $12K.
Copilot for JetBrains gets enterprise MCP allowlists and agent permission locks
GitHub brought MCP allowlists, marketplace locks, managed OTEL, and Autopilot/Bypass locks to Copilot for JetBrains.
Claude Code 2.1.235 adds prompt spellcheck and stops LSP reconnects from busting cache
Claude Code 2.1.235 adds local prompt spellcheck, fixes LSP reconnect cache busts, and lightens background review/fix streams.
Anthropic extends Claude Code’s +50% weekly limits through Aug 31
Claude Code’s +50% weekly-limit promo now runs through Aug 31; 5-hour caps are unchanged and permanence is not guaranteed.
OpenAI pauses deploy-bound RL after Astra may hit Critical cyber
OpenAI paused deploy-bound RL and left its largest frontier run on hold after Astra may meet its Critical cybersecurity threshold.
Claude Code adds a research-preview /design skill for in-session UI artboards
Claude Code’s research-preview /design skill brings Claude Design artboards into CLI and Desktop so agents can mock UI, then implement the chosen option.
Wiz Red Agent exploited a GitHub Actions injection at Snowflake
Wiz’s Red Agent exploited a five-day-old GitHub Actions injection in Snowflake’s .NET connector that GHAS missed, reaching internal Jira credentials.
Hermes Agent v0.20.4 rolls up 74 PRs for Docker and fresh installs
Nous Research tagged Hermes Agent v0.20.4 as a 74-PR stable rollup with skill-install scanning, bot-mode fixes, and SessionDB contention patches.
Alipay launches a merchant agentic commerce platform with MCP tools
Alipay launched a merchant agentic commerce stack that turns pages, products, and workflows into Skills and MCP tools for AI shopping agents.
Hermes Desktop now defaults to named Bot Mode agents
Hermes Desktop now defaults to Bot Mode—named specialist agents with isolated memory and bot-to-bot messaging.
Claude Code 2.1.234 auto-resumes after usage-limit reset
Claude Code 2.1.234 can wait out a claude.ai usage-limit reset, adds a GitLab MR status badge, and hardens leftover NT-namespace file reads.
Cursor Origin early beta is live: a git host built for agent commit volume
Cursor Origin early beta is live on paid plans: native git hosting, GitHub sync, in-editor PRs, and agents on one surface.
SpaceX's $60B Cursor Acquisition Officially Closes, Cementing the Largest Agentic-Coding M&A Deal in History
SpaceX has officially closed its $60B all-stock acquisition of Cursor maker Anysphere, completing the largest agentic coding tool M&A deal in history.
Anthropic Claude Services Suffer Brief Authentication Outage Across Claude Code, Claude.ai, and API
A brief ~30-minute authentication outage hit Claude.ai, the Claude API, and Claude Code on Aug 16, resolved within the hour.
Dario Amodei Publicly Rejects Claim That AI Regulation Concentrates Power
Anthropic's Dario Amodei rebuts claims that AI regulation concentrates power, pointing to California SB 53's threshold-based, pro-competitive design.
Grok 4.6 Lands in GitHub Copilot Across VS Code, Visual Studio, and Copilot CLI
Grok 4.6 lands in GitHub Copilot's model picker across VS Code, Visual Studio, and Copilot CLI, extending xAI's long-running-agent model further into mainstream dev tools.
Cross-Model 'Reasoning Trace Theft' Flaw Lets Attackers Force Plaintext Disclosure of Hidden Chain-of-Thought Across Anthropic, OpenAI, and Google APIs
A cross-provider flaw lets attackers force weaker sibling models to decrypt and leak proprietary chain-of-thought reasoning from Anthropic, OpenAI, and Google APIs.
Writer Launches Palmyra X6 and a Rebuilt Agent Harness, Claims 52% Cheaper Long-Running Agentic AI
Writer's Palmyra X6 and rebuilt Agent harness claim 52% cheaper, 48% faster long-running agentic AI while matching Claude Opus 4.8 on internal evals.
Anthropic's IPO Valuation Reportedly Hinges on a $190-200B 2028 Revenue Forecast
Anthropic's IPO valuation reportedly hinges on an internal $190-200B 2028 revenue forecast, far above its current $47B run rate, per a Reuters exclusive.
Claude Code v2.1.233 Released — GitLab MR Support, Memory Cgroups, and an NTLM Credential-Leak Fix
Claude Code v2.1.233 adds GitLab MR support and memory cgroups, fixes an NTLM credential-leak bug, and disables todo tools by default on newer models.
OpenAI Replaces Chronicle's Screenshot Capture with 'Computer History' Event Logging
OpenAI swaps ChatGPT's screenshot-based Chronicle memory for event-based 'Computer History' logging, aiming to dodge a Microsoft Recall-style backlash.
OpenClaw Team Commits to a Specific Aug 18 Release Date, Details What's Behind the Delay
OpenClaw's team names Tuesday, Aug 18 as its target release date, detailing an onboarding UX overhaul and storage-layer rebuild behind the delay.