Claude dynamic workflows seemed pretty cool and I wanted an open-source version, so I made one. This one improves the test coverage of your entire repo, but you should be able to swap out the prompt and do anything else: https://github.com/OpenHands/software-agent-sdk/pull/3426…

OpenHands/software-agent-sdk

Source: https://github.com/OpenHands/software-agent-sdk

OpenHands Software Agent SDK

Deutsch | Español | français | 日本語 | 한국어 | Português | Русский | 中文

---

The OpenHands Software Agent SDK is a set of Python and REST APIs for building agents that work with code.

You can use the OpenHands Software Agent SDK for:

• One-off tasks, like building a README for your repo
• Routine maintenance tasks, like updating dependencies
• Major tasks that involve multiple agents, like refactors and rewrites

Importantly, agents can either use the local machine as their workspace, or run inside ephemeral workspaces (e.g. in Docker or Kubernetes) using the Agent Server.

You can even use the SDK to build new developer experiences: it’s the engine behind the OpenHands CLI and OpenHands Cloud.

Get started with some examples or check out the docs to learn more.

Quick Start

Here’s what building with the SDK looks like:

import os

from openhands.sdk import LLM, Agent, Conversation, Tool
from openhands.tools.file_editor import FileEditorTool
from openhands.tools.task_tracker import TaskTrackerTool
from openhands.tools.terminal import TerminalTool


llm = LLM(
    model="gpt-5.5",
    api_key=os.getenv("LLM_API_KEY"),
)

agent = Agent(
    llm=llm,
    tools=[
        Tool(name=TerminalTool.name),
        Tool(name=FileEditorTool.name),
        Tool(name=TaskTrackerTool.name),
    ],
)

cwd = os.getcwd()
conversation = Conversation(agent=agent, workspace=cwd)

conversation.send_message("Write 3 facts about the current project into FACTS.txt.")
conversation.run()
print("All done!")

For installation instructions and detailed setup, see the Getting Started Guide. For local development from this repository, run make build to install the workspace dependencies and pre-commit hooks.

Documentation

For detailed documentation, tutorials, and API reference, visit:

https://docs.openhands.dev/sdk

The documentation includes:

• Getting Started Guide - Installation and setup
• Architecture & Core Concepts - Agents, tools, workspaces, and more
• Guides - Hello World, custom tools, MCP, skills, and more
• Agent Server API Reference - REST API reference for the remote agent server

Examples

The examples/ directory contains comprehensive usage examples:

• Standalone SDK (examples/01_standalone_sdk/) - Basic agent usage, custom tools, and skills
• Remote Agent Server (examples/02_remote_agent_server/) - Client-server architecture and WebSocket connections
• GitHub Workflows (examples/03_github_workflows/) - CI/CD integration and automated workflows

Skills for modern package tooling

If you enable public skills with AgentContext(load_public_skills=True), the default OpenHands/extensions marketplace includes, for example, uv and deno skills. Agents can automatically pick up current package-management guidance for repositories that use markers like uv.lock, deno.json, deno.jsonc, or deno.lock.

See examples/01_standalone_sdk/03_activate_skill.py for a minimal example that turns on public skill loading.

Contributing

For development setup, testing, and contribution guidelines, see DEVELOPMENT.md.

Community

• Join Slack - Connect with the OpenHands community
• GitHub Repository - Source code and issues
• Documentation - Complete documentation

Cite

@misc{wang2025openhandssoftwareagentsdk,
      title={The OpenHands Software Agent SDK: A Composable and Extensible Foundation for Production Agents}, 
      author={Xingyao Wang and Simon Rosenberg and Juan Michelini and Calvin Smith and Hoang Tran and Engel Nyst and Rohit Malhotra and Xuhui Zhou and Valerie Chen and Robert Brennan and Graham Neubig},
      year={2025},
      eprint={2511.03690},
      archivePrefix={arXiv},
      primaryClass={cs.SE},
      url={https://arxiv.org/abs/2511.03690}, 
}

Thank You to Our Contributors

Trusted by Engineers at

ClaudeDevs (@ClaudeDevs): New in Claude Code (research preview): dynamic workflows.

Claude writes an orchestration script on the fly, then spins up a large fleet of coordinated subagents in parallel to take on your most complex tasks.

Use the word “workflow” in a prompt to get started.