HKUDS/OpenHarness

Project overview

It bundles the core building blocks an LLM needs to function as a persistent agent—tools, skills, memory, and multi-agent coordination—into lightweight infrastructure that is self-hosted and local-runnable.

Project type

MCP · AI Agent

Use cases

Coding & Development

Deployment

Refer to project documentation

License

MIT

Best for

• Developers who need an LLM-based agent with tool-use, skills, memory, and multi-agent coordination for long-session workflows

Key capabilities

• Previews runtime settings, auth state, skills, and tools without executing the model or tools.
• Includes an MCP client with HTTP transport and auto-reconnect.

Limitations and risks

• On Windows PowerShell, use `openh` instead of `oh` because `oh` can resolve to the built-in `Out-Host` alias.

Getting started

• Install with pip using `pip install openharness-ai`. An install script and an interactive setup wizard are also available; the setup process involves selecting and authenticating with a model provider such as Claude, OpenAI, or Ollama.

Evidence and sources

• GitHub project description: OpenHarness: Open Agent Harness with a Built-in Personal Agent--Ohmo!
• README: openharness/
• README: An **Agent Harness** is the complete infrastructure that wraps around an LLM to make it a functional agent.
• README: 2026-04-01 🎨 v0.1.0 — Initial OpenHarness open-source release
• README: Tests-114_Passing

Source updated at: 2026-07-16T18:10:06Z

Open original GitHub project