QDivZero Harness
Your models, files and agents. In one Mac app.
An idea needs more than an answer. QDivZero Harness connects AI to your project to research, write code and create documents. Follow the agent’s activity, review its changes and open the result without leaving the conversation.
macOS 13 or later · Apple Silicon · v0.19.0
The app is free and open source. You choose the provider and how to pay for its models.
01 /From model to action
A model can explain how to solve a task. A harness gives it access to the environment where the work needs to happen. QDivZero Harness connects the conversation to your files, terminal and web research to make progress on a real project.
Ask it to explore a repository, prepare a document or research a proposal. Its tool use and activity appear in the conversation. The inspector brings files, context and changes together so you can follow the work.
01 / The right model
Bring together connections to QDivZero, OpenAI, Anthropic, DeepSeek and other providers. Each keeps its own catalog and credentials. You can also connect Ollama and use local models in the same app.
02 / Tools that act
Harness can read and edit files, apply patches, use the terminal and research with search and a browser. Your project’s AGENTS.md instructions and skills provide guidance for working in your environment.
03 / Results you can review
Documents, images and PDFs generated by the agent appear as deliverables in the conversation. Preview them in the app and use Files, Context and Changes to review how it reached the result.
02 /Choose how you work with AI
When comparing AI harnesses, think about how you will work each day. QDivZero Harness brings multiple providers, tools and deliverables into a native Mac app written in Rust. Use it to develop software, research a topic or prepare documents.
Its approach combines model choice with a visual workspace for following context, reviewing changes and collecting results. Memory, delegated agents and MCP connections let you adapt it to each project.
Connect MCP servers per project and load skills when a task calls for them. Harness reads your AGENTS.md instructions to follow the repository’s rules.
Search earlier conversations through the memory index. When you want to save a fact for future sessions, global memory asks for your approval before keeping it.
Assign a task to an agent with its own connection and model. Review its activity and tool use from the main conversation.
Remote Agent connects Telegram to the same live conversation on your Mac. Access controls and approvals let you continue working remotely.
Starts with a minimal terminal coding harness and lets you extend its behavior through extensions and skills.
Offers an open-source coding agent across terminal, desktop and IDE, with access to different providers.
Offers a Claude development workflow across terminal, IDE, desktop and web, connected to its coding tools.
With QDivZero Harness, your Mac workspace is the starting point. Choose the connection, bring the context together and follow the agent through to the file or change you need.
03 /Your first project with Harness
Download the app for a Mac with Apple Silicon and macOS 13 or later. The version 0.19.0 installer is signed and notarized. Get started in the app without setting up a development environment.
Open the DMG, drag QDivZero Harness to Applications and launch it from there. You now have a workspace for your conversations and projects.
Add a connection in Settings and select a model from its catalog. Use your provider’s API key or sign-in flow. Local Ollama on your Mac does not require an API key.
Choose your workspace folder and explain what you want to achieve. Follow the agent’s tool use, review changes and open the documents it generates.
Download the installer from the official QDivZero Harness release on GitHub.
Download SHA256SUMS.txt and save it alongside the DMG. Open Terminal in that folder and run this command to check that the file matches the published download.
shasum -a 256 -c SHA256SUMS.txtConnect QDivZero, OpenAI, Anthropic, DeepSeek, Groq, Ollama and Ollama Cloud. You can also add an OpenAI-compatible API or use the Codex and GitHub Copilot sign-in integrations.
Models and access conditions depend on each connection. Running a local model requires sufficient resources on your Mac.
04 /A workspace you can make your own
QDivZero Harness is free and released under the MIT license. Download it, inspect how it works and adapt its code to your needs. You do not need a Harness subscription to get started.
You choose which models to connect. If you use a provider API or subscription, its usage is paid under that service’s terms. With Ollama, you can work with local models according to your Mac’s capacity. Downloading Harness does not include third-party credits or subscriptions.
05 /Your questions, answered
What you need to choose QDivZero Harness, install it and get started with your own AI connections.
Alongside a conversation, you can give a model tools and project context to carry out tasks. Harness brings files, terminal, browser, skills, MCP, memory and delegated agents together. Follow its activity and review the files or documents it generates inside the app.
The app is free and open source under the MIT license. You do not need a Harness subscription. APIs and subscriptions from connected providers are paid separately under their terms. The app does not include inference credits for those services.
The version 0.19.0 installer is available for Macs with Apple Silicon and macOS 13 or later. This release does not include installers for Intel Macs, Windows or Linux.
Yes. Connect QDivZero, OpenAI, Anthropic, DeepSeek, Groq, Ollama, Ollama Cloud and OpenAI-compatible APIs. Codex and GitHub Copilot sign-in integrations are also available. Each connection shows the models available to your account with that provider.
History and work data are kept in a local SQLite database. Credentials are stored in the app’s encrypted vault backed by the macOS Keychain. When you use remote models or services, the data required to handle the request is sent to them under the selected provider’s terms.
Download the version 0.19.0 DMG from GitHub, open it and drag the app to Applications. Add a connection in Settings and select a model. Then open your workspace folder and describe a task. You can verify the DMG with SHA256SUMS.txt by following the steps on this page.
Open your project, connect your model and let QDivZero Harness help you take it through to a result.
Start with Harness for freeMac with Apple Silicon · macOS 13 or later · MIT license