# Agentle

A Python framework for building and deploying production AI agents with a small amount of code, including type-safe structured outputs, built-in observability, automatic REST API generation, and Model Context Protocol integration.

Agentle lets an agent be defined in a few lines with a name, a model, a generation provider and instructions, then run directly or exposed as a REST API with automatically generated documentation. It supports agent pipelines that chain agents together for multi-step workflows, agent teams with dynamic routing across several agents, and MCP integration for connecting to external tools.

Production features include built-in observability through Langfuse, type-safe outputs backed by Pydantic, document caching to avoid repeatedly parsing the same large document or re-calling a paid API for it, with in-memory or Redis-backed caching for sharing cached documents across processes, and a Streamlit integration for building a chat interface around an agent in a few lines.

It requires Python 3.13 or later, installs with pip, and is licensed under MIT.

Website: https://paragon-intelligence.github.io/agentle/