Anthropic ships Claude Fable 5.1 and Mythos 5.1, JetBrains debuts Air and Console, and Adobe AEM adds an MCP server as agent tooling expands

Anthropic rolls out Claude Fable 5.1 and Claude Mythos 5.1 for coding and knowledge work

Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1 on 2026-09-23, describing them as "the world's most advanced models for coding and knowledge work" in its launch post. The post also flags a research focus tied to the release, framing both models as successors in the Fable and Mythos lines that practitioners building with Anthropic models will be expected to evaluate. For a developer choosing a model today, the practical effect is that benchmark claims, context-window handling, and tool-use reliability around agentic workflows will need re-checking against the 5.1 versions rather than previous Fable or Mythos builds.

JetBrains launches Air, an agent-building system, alongside a new enterprise Console

JetBrains used 2026-09-23 to announce two complementary products. JetBrains Air is described as "one system for building software with agents" that "works with the agents you already use and any ACP-compatible agent," signalling an openness to third-party agent runtimes rather than a JetBrains-only agent stack. The JetBrains Console, announced on the JetBrains company blog, "provides enhanced AI management and analytics for organizations, including new capabilities to manage" AI usage at an admin level.

Together the two releases split JetBrains' AI pitch: Air is the build surface for agentic coding, while the Console is the governance and observability layer for teams rolling those agents out. Teams that already standardise on JetBrains IDEs will need to decide whether to anchor agentic work to Air or keep using their current ACP-compatible tool.

Adobe Experience Manager exposes an MCP server for AI-powered IDEs and chat tools

Adobe Experience Manager as a Cloud Service documented an AEM Model Context Protocol (MCP) Server that "brings AEM into your preferred AI-powered IDE or chat-based application, to streamline and accelerate your AEM work." In practical terms, an MCP endpoint for AEM means editors and developers who live in tools like Claude, ChatGPT clients, or AI-enabled IDEs can read and act on AEM content and configuration through a standardised protocol instead of a one-off connector. For front-end and content-tooling teams this lowers the friction of integrating AEM into agentic workflows, and pairs naturally with the ACP-compatible agents JetBrains Air is targeting.

Datadog Incident Management adds Zoom meeting creation for live incidents

Datadog's incident management documentation gained an integration page explaining how to "quickly create Zoom meetings to collaborate with your team on active incidents" by connecting Zoom to Datadog. The setup is straightforward — link the two accounts and trigger a Zoom room from inside a Datadog incident — which matters for on-call rotations that already pay for both products and want fewer context switches during a page. SREs and platform teams evaluating incident tooling on 2026-09-23 can add this to the list of native bridges Datadog offers versus building a webhook themselves.

A new GitHub repository formalises skills for AI coding agents working on .NET

NET package" workflows. NET tooling is being packaged into discrete, callable skills that an AI coding agent can invoke the same way a developer would run `dotnet build` or manage packages. For teams adopting agentic coding on the Microsoft stack, this repository is worth watching as a signal of how AI agent skills will be distributed, versioned, and maintained alongside the SDK itself.

Readers evaluating AI assistants for a C# codebase can compare how well each agent consumes skills from this repository before committing to a platform.

What to check next

1 versus whatever you currently use for coding tasks), the agent runtime (whether JetBrains Air or another ACP-compatible system is the right host), and the protocol surface (the new AEM MCP server as a precedent for connecting more of your CMS or data stack to AI clients). For incident response, the Datadog–Zoom integration is a low-cost bridge to enable now if both tools are already licensed. NET agent skills, since its contents will likely shape how future AI assistants handle MSBuild failures and NuGet operations.

No deadline is given in the evidence for any of these releases; check each vendor's changelog directly for version notes, deprecations, and migration guidance.