Mirascope

Mirascope is an open-source Python library that takes a decorator-based approach to building LLM-powered applications. Instead of wrapping interactions in heavy frameworks, it gives developers direct control over prompts, model selection, and execution flow using familiar Python patterns. The core philosophy is minimal abstraction—you compose LLM calls, tools, versioning, and tracing directly in your code, retaining full ownership of the logic. It supports multiple providers including OpenAI, Anthropic, and Google through a unified interface, making it easy to switch or mix models. Key features include automatic versioning with @ops.version, per-call tracing and cost tracking via @ops.trace, built-in tool calling with @llm.tool, agent loops that execute tools and resume responses, streaming support, and thinking/reasoning capabilities via the include_thoughts parameter. The library is designed for production use, with OpenTelemetry integration for observability and structured output through Python type hints. Its lightweight nature means no heavy dependencies or framework lock-in. Mirascope is ideal for developers who want fine-grained control over their LLM interactions and need built-in observability without sacrificing code clarity. It's particularly well-suited for teams building custom agents, multi-provider systems, or applications that require rigorous cost tracking and versioning. The latest release showcases a mature API that handles both simple calls and complex agent loops with equal elegance. Compared to heavier alternatives like LangChain, Mirascope gives you less out-of-the-box but more control. It's not a framework you build on top of—it's a set of tools you integrate into your existing Python code. This makes it a strong choice for experienced developers who value transparency and want to avoid the debugging headaches that come with high-level abstractions.