Simon Willison released version 0.27 of llm-anthropic, the Anthropic plugin for his LLM command line tool, to restore compatibility after Anthropic shipped version 1.0 of its official Python library, according to Willison's blog.
Anthropic's v1.0 release replaced the HTTP library the SDK depends on, moving from httpx to httpx2, Willison wrote. OpenAI made the same underlying change in its own Python library two weeks earlier, in its v3.0.0 release.
Willison said he used Claude Code, prompting Anthropic's Fable 5 model to read the official migration guide and resolve the test failures the dependency swap caused. The fix shipped as pull request 84 on the llm-anthropic GitHub repository.
The httpx2 swap is a breaking change for any code that touches the Anthropic Python library's underlying transport, timeouts or connection pooling directly. Anthropic's migration guide is required reading before upgrading, even when the public API on top looks unchanged.
Any team with code pinned to the Anthropic or OpenAI Python SDKs should check dependency constraints before the next upgrade. A major version bump that swaps the HTTP client under the hood can break tooling built on top even when nothing in the documented interface changed.