Mirascope

// Overview

Mirascope is a Python toolkit that lets you write LLM calls as regular, typed functions. You decorate a function with @llm.call, return your prompt as the function body, and Mirascope handles sending it to the model and giving you back a response object.

It gives you one unified interface across frontier LLM providers, so the same function shape works whether you target Anthropic, OpenAI, or others. You can also ask for structured output by passing a Pydantic model, and the result is parsed back into that typed object.

It fits the prompt-programming corner of LLM orchestration: instead of stitching together prompt strings and raw API clients by hand, you keep prompts, calls, tools, and parsing inside normal Python functions. It suits developers who want type hints and Pydantic models rather than a heavier framework.

// What it does

• Define an LLM call by decorating a Python function with @llm.call and returning the prompt from the function body
• One unified interface to use any frontier LLM, switching providers by changing the model string
• Structured output via Pydantic: pass format=YourModel and call .parse() to get a typed object back
• Tool calling with @llm.tool, including a resume/execute_tools loop for agent-style multi-step runs
• Supports streaming, async, and multi-turn conversations (see the full documentation)
• Lightweight and typed, building on Pydantic models for inputs and outputs

// Getting started

Install Mirascope, then write your first call as a decorated Python function. All snippets below are taken directly from the project README.

uv add "mirascope[all]"

from mirascope import llm

@llm.call("anthropic/claude-sonnet-4-5")
def recommend_book(genre: str):
    return f"Recommend a {genre} book."

response = recommend_book("fantasy")
print(response.text())

from pydantic import BaseModel
from mirascope import llm

class Book(BaseModel):
    title: str
    author: str

@llm.call("anthropic/claude-sonnet-4-5", format=Book)
def recommend_book(genre: str):
    return f"Recommend a {genre} book."

book = recommend_book("fantasy").parse()
print(f"{book.title} by {book.author}")

from pydantic import BaseModel
from mirascope import llm

class Book(BaseModel):
    title: str
    author: str

@llm.tool
def get_available_books(genre: str) -> list[Book]:
    """Get available books in the library by genre."""
    return [Book(title="The Name of the Wind", author="Patrick Rothfuss")]

@llm.call("anthropic/claude-sonnet-4-5", tools=[get_available_books], format=Book)
def librarian(request: str):
    return f"You are a librarian. Help the user: {request}"

response = librarian("I want a fantasy book")
while response.tool_calls:
    response = response.resume(response.execute_tools())
book = response.parse()
print(f"Recommending: {book.title} by {book.author}")

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Writing LLM calls as typed Python functions instead of hand-managing prompt strings and raw API clients
• Extracting structured data from text into Pydantic models with validated, typed output
• Switching between LLM providers behind one interface without rewriting call code
• Building tool-using agents that loop through tool calls and resume until they produce a final answer

// How Mirascope compares

Mirascope alongside other open-source prompt programming tools AI/TLDR tracks, ranked by GitHub stars.