Acorn

LLM agent framework with structured I/O. Built on Pydantic for schemas and LiteLLM for multi-provider access.

What is Acorn?

Acorn makes it easy to build LLM agents that take typed inputs and return typed outputs. Define a Module, give it tools, and invoke it like a function.

import asyncio
from acorn import Module
from pydantic import BaseModel, Field

class Output(BaseModel):
    sentiment: str = Field(description="positive, negative, or neutral")
    confidence: float = Field(description="Confidence score 0-1")

class SentimentAnalyzer(Module):
    """Analyze sentiment of text."""
    final_output = Output

# Async
async def main():
    analyzer = SentimentAnalyzer()
    result = await analyzer(text="I love this product!")
    print(result.sentiment)  # "positive"

asyncio.run(main())

For synchronous code, use .run():

# Sync
analyzer = SentimentAnalyzer()
result = analyzer.run(text="I love this product!")
print(result.sentiment)  # "positive"

Features

• Structured I/O - Pydantic models for inputs and outputs with automatic validation
• Multi-provider - Works with any LLM via LiteLLM (Anthropic, OpenAI, Google, Ollama, etc.)
• Tool calling - Define tools with @tool decorator, schemas auto-generated from type hints
• Async tools - Tools can be async def and execute concurrently when called together
• Agentic loops - Multi-step ReAct loops with configurable step limits
• Branching - Spawn sub-agents that inherit context and return structured results
• Services - Group related tools with shared config, auth, and lifecycle management
• Tool Discovery - Search-based tool discovery for agents with many tools
• Streaming - Stream text and partial structured outputs
• Callbacks - Inspect and modify execution at every step

Quick links

• Getting Started - Install and build your first agents
• Module - Full reference for the Module class
• Branching - Sub-agents and parallel processing
• Services - Tool collections, lifecycle, and built-in services
• Tool Discovery - Search-based tool discovery
• GitHub