stirrup

What it is and what it does

Stirrup is a lightweight agent framework designed to work with language models rather than impose rigid workflows on them. It provides a foundation for building autonomous agents that can execute code (locally, in Docker, or E2B sandboxes), search and fetch web content, call external tools via MCP servers, process files, and incorporate human feedback. The framework includes built-in context management to handle long conversations, multimodal support for images and video, and a flexible tool system based on Pydantic schemas.

The package ships with sensible defaults—local code execution and web search tools—but is fully customizable. You can use it as an installed package or fork it as a template for building specialized agents. It supports multiple LLM providers through OpenAI-compatible APIs, LiteLLM, or custom clients, and includes a skills system for modular, domain-specific instruction packages. The 11 runtime dependencies cover HTTP requests, image and video processing, schema validation, and retry logic.

Use it for

• Build an autonomous research agent that searches the web, fetches articles, and synthesizes findings into a report
• Create a code-generation agent that writes, tests, and refines scripts based on user requirements
• Develop a data analysis agent that processes files, executes analysis code, and generates visualizations
• Set up a multi-tool agent that integrates with external services via MCP to automate complex workflows
• Prototype a specialized agent by cloning the repository and customizing the core loop and tool set

Worth the install?

AI-flagged interpretation of the facts on this page. Verify before relying on it.

Install

stirrup on PyPI

Before you install

Low friction installation with a pure Python wheel. Active maintenance—last commit 2026-08-05, released 2026-08-04—and 548 repository stars suggest ongoing development. Requires Python 3.12 or later.

Requires Python 3.12 or later. API key environment variable (OPENROUTER_API_KEY or equivalent for your provider) must be set before running.

License in practice

MIT License permits unrestricted use, modification, and redistribution with minimal restrictions, making it suitable for both commercial and open-source projects.

Quickstart

pip install stirrup

import asyncio
from stirrup import Agent
from stirrup.clients.chat_completions_client import ChatCompletionsClient

async def main():
    client = ChatCompletionsClient(
        base_url="https://openrouter.ai/api/v1",
        model="anthropic/claude-opus-5",
        max_tokens=8_192,
        context_window_tokens=1_000_000,
    )
    agent = Agent(client=client, name="agent", max_turns=15)
    async with agent.session(output_dir="./output") as session:
        finish_params, history, metadata = await session.run("Your task here")

asyncio.run(main())

Package facts

Evidence: stirrup-0.2.0-py3-none-any.whl

Tags

More Python Modules packages

See also ai-parrot · strands-agents · fast-agent-mcp · langchain · npmai-agents · smolagents · PraisonAI · openai-agents · livekit-plugins-anthropic · deepagents

Further reading