Stirrup

Stirrup is an open-source Python framework for building AI agents, designed to be lightweight and flexible. Instead of imposing rigid workflows, it gets out of the model's way, letting it choose its own approach to completing tasks — similar to Claude Code. The framework includes best practices and essential tools built-in, such as web search, code execution, an MCP client, and document handling. It's fully customizable, usable as a package or as a starting template to build fully customized agents. Stirrup is ideal for developers who want to build agentic applications without being locked into rigid workflows. It supports multiple LLM providers via OpenAI-compatible APIs, LiteLLM (Anthropic, Google, etc.), or custom clients. The framework includes context management that automatically summarizes conversation history when approaching context limits, and multimodal support for images, video, and audio. What makes Stirrup different is its philosophy: working with the model, not against it. Many agent frameworks impose fixed workflows that degrade results, but Stirrup lets the model orchestrate its own approach. It incorporates best practices from leading agents like Claude Code and Codex, and offers a flexible tool execution system with a generic Tool class for easy extension. The Skills system allows modular, domain-specific instruction packages to extend agent capabilities. Stirrup can be installed via pip or uv, with optional extras for LiteLLM, Docker, E2B, MCP, and browser support. It uses a session-based execution model with built-in logging and file output handling. Compared to heavier frameworks like LangChain or CrewAI, Stirrup is much more minimal and lets the model drive decisions, making it a great choice for developers who want full control without unnecessary abstraction.