mirascope

Mirascope lets you use any frontier LLM through one unified interface. In Python you decorate a function with `@llm.call("anthropic/claude-sonnet-4-5")` and call it; the model is named by a provider/model string. The README shows three steps: a basic call, structured output by passing a Pydantic model as `format=` and reading it with `.parse()`, and an agent loop where functions marked `@llm.tool` are passed as `tools=` and the response is resumed with `response.resume(response.execute_tools())` while it has tool calls. Streaming, async and multi-turn conversations are covered in the docs.

Install with `uv add "mirascope[all]"`. The repository is a monorepo with a Python implementation (with examples), a TypeScript implementation, a `cloud/` full-stack application (React frontend and Cloudflare Workers backend) and unified docs. It follows Semantic Versioning, and CI runs spell-check, Python lint and type checks, TypeScript lint and a docs build.

Licensing is split: the code is MIT unless otherwise specified, but the `cloud/` directory is under a proprietary license, and subdirectories may carry their own LICENSE files. It was last pushed on 2026-10-12. This review is based on the README and repository data; we have not run it ourselves.