llm-anthropic 0.27
Simon Willison’s Weblog Simon Willison ● Covered by 3 sources
llm-anthropic 0.27 is out, mainly to keep up with Anthropic’s new Python library. It also shows OpenAI and Anthropic are both making the same httpx2 switch.
Based on reporting by Simon Willison’s Weblog, Simon Willison — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
llm-anthropic 0.27 landed on 24 August 2026 as a compatibility release for Anthropic’s Claude-facing plugin in Simon Willison’s LLM tool.
The main reason for the bump is simple: Anthropic’s newly released v1.0.0 Python library changed from httpx to httpx2, and this plugin had to follow along. The release is mostly about making sure LLM access to Anthropic models keeps working cleanly with that new client.
OpenAI, for its part, made the same move in v3.0.0 two weeks earlier. So this isn’t just one vendor’s housekeeping. It’s a shared shift in the plumbing that developers lean on, and those shifts tend to matter most when they break at the worst possible moment.
Willison says Anthropic provides a migration guide for the 1.0 upgrade, and he used that directly. He prompted Fable 5 in Claude Code with a very specific job: upgrade to anthropic>=1, read the migration guide, and get the tests passing. The result was a pull request, which is exactly the sort of boring, practical outcome you want from a dependency upgrade.
Posted the same day at 4:27 pm, the release reads less like a big feature drop and more like careful maintenance. That’s usually the right kind of software news: unglamorous, necessary, and one step ahead of the next broken build.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI tooling that actually deserves attention: the unsexy dependency work. Anyone selling “agents” while ignoring library churn is peddling smoke, because the real product is usually test fixes and migration guides. Open models or closed ones, the useful stack is the one that survives an upgrade without drama.
Read more about this at: Simon Willison’s Weblog
