open-science
Open Science is an open-source, local-first, model-agnostic AI research workbench for scientific discovery.
What is it?
What it is
Open Science is an open-source, local-first, model-agnostic AI research workbench for scientific discovery.
Why it exists
To address fragmented research workflows by providing a single inspectable desktop workspace where AI agents can execute tasks, generate reproducible artifacts, and maintain full provenance, thus preventing context loss and separating answers from their production evidence.
Who should use it
researchersscientistsdata analystsacademicsprofessionals in bioinformatics or medical research
Who should avoid it
non-technical userscasual users without setup requirementsthose needing cloud-only solutions
How it works
A quick walkthrough in plain English
How open-science works
Step 1 of 3
You interact with it
Open open-science, send a request, or connect it to your stack.
Features
Advantages
- Ensures reproducibility through detailed execution history and environment evidence
- Reduces context loss by unifying chat, notebooks, and file management in one workspace
- Provides high flexibility with multiple model provider options (Cloud, Custom Gateway, Subscriptions)
- Maintains data privacy with local-first storage and local logs
- Extensible architecture via custom skills and MCP connectors
- Open-source (Apache-2.0) with no seat licensing fees
Disadvantages
- Requires manual configuration/authentication for various model providers
- External data flow still occurs during web searches and connector calls
- Complexity in managing environment dependencies for Python/R execution
- Requires user oversight to prevent misuse of sensitive local paths or data
Installation
native
1. Download the latest release installer for your OS (macOS DMG, Windows x64 installer, Linux AppImage or Debian package). 2. Run the installer to install Open Science. 3. On first launch, complete the Prepare environment and Model provider setup steps. 4. Create a new project, start a session, attach files, describe your task, select a model, and send the task to begin research.
FAQ
What should I do the first time I open Open Science?
Complete the 'Prepare environment' and 'Model provider' steps. You must fix any rows marked 'Action needed' and ensure the model connection test passes before you can continue.
Do I need an API Key to use the application?
Not necessarily. You can reuse existing subscriptions by signing in via a Claude setup-token or a ChatGPT/Codex subscription on the Codex backend. However, built-in cloud providers and custom gateways require their own API keys.
How can I verify the origin and reproducibility of a generated result?
You can use the 'Provenance' view on any generated artifact. This allows you to inspect the producer code, execution history, inputs, environment inventory, and the specific conversation branch that produced the result.
Can I explore different research hypotheses without losing my previous work?
Yes. You can edit a completed user message to create a new message branch. This allows you to explore alternative research paths while keeping the original conversation and its subsequent turns intact.
Is my research data stored on the cloud or my local machine?
By default, projects, sessions, files, settings, and credentials are stored locally on your computer. Note that content sent to model providers or web searches is transmitted to those external services.
Does Open Science offer a command-line interface (CLI)?
Yes. You can install the CLI via Settings → General. The CLI allows you to control the local service and automate research tasks, such as creating projects or downloading artifacts, without opening the desktop UI.
Featured in Videos
YouTube tutorials and walkthroughs for open-science
Alternatives
Similar projects ranked by category, topics, and text overlap.
