What is open-science?

Open Science is a free desktop application that gives researchers an AI assistant for scientific work. Instead of writing code or scripts yourself, you describe a task in plain language, such as analyzing a dataset or producing a report, and an AI agent inside the app reads your files, runs code, searches the web, calls scientific data connectors, and returns reports, tables, and figures, along with a record of every step it took. The app works as a project based workspace. You create a project, open a session, describe your goal along with the input data and any constraints, attach the relevant files, pick a verified AI model, and choose how much you want to approve each action the agent takes before it runs. As the agent works, you can inspect exactly which tools it used, which files it read or edited, and which searches or connectors it called, and you can open the generated results directly in a preview panel inside the app. Open Science is described as model agnostic, meaning you are not locked into one AI provider. During first time setup you can connect a built in model provider, an Anthropic compatible custom gateway, or reuse an existing subscription login such as a Claude Code login or a ChatGPT or Codex subscription, without needing to enter a separate API key. The app can also install its own managed runtime automatically if needed, without requiring you to separately install Node.js or have administrator access. The project includes previews for common scientific file types such as data tables, documents, images, source code, molecular structures and reactions, and notebook history, and it ships with a growing catalog of file based research skills that can be extended with custom skill packages. Installers are available for macOS, Windows, and Linux from the project's releases page, and the project is released under the Apache 2.0 license.