New Claude Code Update is INSANE!
Anthropic has released a Claude API skill bundled inside Claude Code that automates API migrations, onboards developers to managed agents, and stays current as the API evolves. The skill can migrate entire projects to new models like Opus 4.7 in one command, handling breaking changes automatically. It is open source and already integrated into tools like Code Rabbit, JetBrains, and Warp.
Summary
The video covers a newly official Claude API skill shipped inside Claude Code that automates several previously time-consuming developer tasks. The central feature is the /claude ap migrate command, which scans an entire project, finds every Claude API call, and updates model names, parameters, beta headers, and effort settings in a single guided pass — a process that previously could take developers days or even a full sprint.
The skill performs three core jobs. First, it teaches Claude how to write Claude API code correctly, covering features like prompt caching, tool use, batch processing, streaming, memory, citations, and compaction, staying current as the API evolves. Second, it handles model migrations, such as updating projects to Opus 4.7. This includes switching from fixed token budgets for thinking to the new adaptive thinking system, removing deprecated sampling parameters like temperature, top_p, and top_k (which now return errors on 4.7), flagging token count differences that could cause overflow, recommending the new x-high effort level, and catching silent behavioral changes like thinking summarization that affect user experience.
Third, the skill onboards developers to Anthropic's managed agents platform, where Anthropic hosts the agent container and runs the agent loop. Developers only define the agent once — model, system prompt, tools, MCP services — and reference it by session ID. The skill runs an interview-style setup and scaffolds working code in Python, TypeScript, Java, Go, Ruby, PHP, or Curl, including proper beta headers, streaming event handling, tool confirmation handling, and lossless reconnect logic.
The skill also automates prompt caching optimization, reviewing code to suggest correct cache breakpoint placement and auditing for silent cache invalidators. It uses progressive disclosure, loading only the documentation relevant to the current task rather than dumping the entire manual into context.
The skill is open source on GitHub at anthropic/skills and can be bundled by any coding agent with approximately 20 lines of CI configuration. It is already live in Code Rabbit, JetBrains, Juni, Resolve AI, and Warp. The presenter frames this as a coordinated rollout of a single source of truth across the entire developer ecosystem, replacing the fragmented landscape of outdated blog posts with one living document that updates as the API does.
Key Insights
- The speaker explains that sampling parameters like temperature, top_p, and top_k have been completely removed on Opus 4.7 and now return errors, meaning any project migrated without the skill would silently break in production.
- The speaker argues that the core problem with most coding agents is that they go stale — trained on old docs and deprecated parameter names — and the Claude API skill solves this by loading the latest documentation every time rather than relying on training data.
- The speaker describes Anthropic's managed agents platform as a shift where developers no longer build or host the agent loop at all — they write a config once, start a session, and Anthropic runs the container with bash, file ops, code execution, and web search built in.
- The speaker notes that the skill uses progressive disclosure, loading only the portion of documentation relevant to the current task — tool use docs for tool use, streaming docs for streaming — rather than injecting the full manual into the context window.
- The speaker points out that the skill is open source on GitHub and any coding agent can bundle it with about 20 lines of CI, meaning the same up-to-date guidance will be available to developers across every IDE and coding tool simultaneously.
Topics
Transcript
[0:00] New Claude code update is insane. You type one command /claude ap migrate and Claude code goes through your entire project, finds every place you call the Claude API, and updates the whole thing for you. Models, parameters, beta headers, effort settings, done. This used to take a developer days, whole sprint sometimes, reading docs, hunting for breaking changes, fixing one file, breaking another. Now it happens in one guided pass. And here's the part most people missed. This skill isn't a plugin you install. It's bundled. Ships inside [0:30] Claude code by default. Every single user has it right now. You don't even need to know it exists for Claude to use it. That changes the game. So…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Julian Goldie SEO
NEW Nvidia Autonomous AI is WILD!! 🤯
Nvidia announced Nemo Clo, a new autonomous AI agent system that operates independently without continuous prompting. Powered by Nemotron 3 Ultra (a 550 billion parameter model), the system is five times faster and cheaper than previous versions, with OpenShell providing secure sandboxed execution.
Laguna XS 2.1: New FREE + Opensource Local AI!
Julian reviews Laguna XS 2.1, a new free open-source local AI coding model from Poolside that performs comparably to Qwen 3.6 and outperforms Claude Haiku on benchmarks. He demonstrates its practical capabilities by building landing pages and functional apps, highlighting its speed, offline functionality, and multiple deployment options through local setup, Claude Code, or OpenRouter's free API.
How to Run Hermes FREE Forever!
The video demonstrates how to run the Hermes AI agent for free using Gemma 4, a local open-source model from Google, with significant speed improvements through MLX optimization. The setup works on Apple Silicon Macs or via free APIs on Open Router, enabling autonomous agents to work offline and privately without subscription costs.
This NEW Chinese AI is INSANE! (FREE + Open Source!)
Long Cap 2.0 is a new open-source Chinese AI model from a food delivery app company that offers 1 million tokens of free context memory, beats GPT-4.5 on SWE bench pro benchmarks, and uses efficient parameter activation to reduce computational overhead while maintaining high performance.
Claude Code is now FREE: Here’s how…
Google's new Gemma 4 model running on Ollama is 90% faster on Apple Silicon, enabling free Claude Code usage locally without token costs. The setup requires three simple steps: downloading Ollama, Gemma 4, and installing into Claude Code, with alternatives available via OpenRouter API for non-Mac users.