If you’ve been using Claude Fable 5 through Anthropic’s extended promotional period, tonight’s the deadline: July 12, 2026, at 11:59:59 PM PT. After that cutoff, Fable 5 access shifts from your included subscription limits to usage credits billed separately — at $10 per million input tokens and $50 per million output tokens. For casual users, this is a budgeting question. For teams running Fable 5 in automated pipelines or agentic workflows, this could mean a significant and unexpected cost jump starting tomorrow if you haven’t already planned for it. ...

Every time you use an AI coding agent, it reads files on your computer. It opens your source code, scans your configuration files, and potentially encounters secrets — API keys, tokens, SSH credentials — that live in your project directories. The agent processes all of this and sends relevant portions to a remote model. Most developers just… trust that process. But what if you didn’t have to? Confessor is a new open-source tool from ninjahawk on GitHub that answers one specific question: What did Claude Code actually do on my computer during that session? ...

The Model Context Protocol is about to go through its biggest change since launch — and the beta SDKs for that change are available right now. If you’re building MCP servers or clients, now is the time to understand what’s coming and start testing before the spec goes final on July 28, 2026. The short version: MCP is going stateless. The initialize handshake is gone. Session pinning is gone. And there are real breaking changes in both the Python and TypeScript SDKs that require migration work before the stable release. ...

Voice AI has always come with an awkward trade-off: the faster and more natural it feels, the less capable it is at complex tasks — and vice versa. The systems best at reasoning tend to be slow and turn-based. The systems that feel conversational struggle with anything that requires real thought. OpenAI launched GPT-Live-1 on July 8, 2026 with a clear goal: break that trade-off. The result is a voice AI that can hold a real conversation while silently routing your hard questions to a much more powerful model working in the background. ...

If you’ve ever wished your AI agent could just know things — not because you told it to remember something, but because it proactively built up its own working knowledge from your data — LangChain may have just shipped the tool you didn’t know you needed. OpenWiki Brains is a new open-source framework that gives AI agents proactive memory. Not reactive memory. Not “store this conversation snippet for later.” Proactive memory that actively reaches out to your connected data sources, fetches relevant context, and builds a living, self-updating wiki — without you having to prompt it to remember anything. ...

If you’ve been running Claude Code for a while, your workspace configuration has probably accumulated some cruft. Skills you installed and forgot. MCP servers you added for one project. A CLAUDE.md file that’s grown into a multi-page manifesto. The /checkup command — also known as /doctor, the underlying alias — was added in v2.1.205 to address exactly this. It’s an interactive workspace audit that can find problems and, with your confirmation, fix them. ...

If you’ve ever caught yourself flipping between your browser and Claude Code to paste in documentation, check a UI change, or verify that the thing you just built actually works — that workflow just became obsolete. Anthropic has shipped a native in-app browser directly inside Claude Code’s desktop app, and it’s a bigger deal than the announcement made it sound. What Just Landed Starting with Claude Code desktop v2.1.202 (rolled out across July 6–10, 2026), the app now includes a full built-in browser pane. This isn’t a stripped-down webview or a limited preview mode — it’s a first-class browsing environment inside the coding environment itself. ...

GitHub’s June 2026 release cycle for Copilot in VS Code — covering versions v1.123 through v1.127 — landed a collection of agentic features that shift Copilot meaningfully toward parallel, browser-capable, cost-conscious agent work. These changes were announced on July 8 via GitHub’s official changelog. If you’ve been tracking Copilot’s evolution from code completion to full agent, this is one of the more substantive monthly updates in that arc. Agentic Browser Tools Are Now Generally Available The headline feature: agentic browser tools for Copilot in VS Code have exited preview and are now generally available, enabled by default. Agents can now navigate web pages, inspect content, capture screenshots, and validate web applications directly within VS Code — without switching to a separate browser. ...

OpenClaw’s beta channel keeps moving fast. The 7.1 series has been shipping steadily since the cycle opened, and beta.5 — tagged July 11 at commit b6387af — brings some of the most user-facing changes yet. If you’re on the beta channel, here’s what’s in this release. If you’re on stable (currently 2026.6.x), this is a preview of where the platform is heading. Crestodian: Conversational AI-Guided Onboarding New users now get Crestodian — an AI-guided onboarding experience that walks through initial setup conversationally across CLI, web, and macOS. Instead of reading docs and running a series of manual config commands, Crestodian asks questions, understands your use case, and guides setup accordingly. ...

Security researcher Chinmohan Nayak has published a detailed breakdown of how three separate flaws in OpenClaw — all patched in version 2026.6.6 — can be chained together to achieve unauthenticated remote code execution on the host machine, starting from nothing more than an external WhatsApp message. If you’re running an older version of OpenClaw with WhatsApp connected, this is a drop-everything update. The Attack Chain The research identifies three distinct vulnerabilities: ...