Skip to content

Get Started with ScienceX ​

ScienceX is an open-source, local-first research twin workbench. For a first visit, read the product position and capability boundaries, then run the three-minute teaching case to complete a digital experiment twin path from a research object and versioned data to analysis artifacts and replay verification.

Choose your path ​

Your goalStart here
Understand the ScienceX product positionResearch Twin Workbench
Use ScienceX directlyDownload the desktop app
Understand the research workflowRun the three-minute teaching case
Run the terminal app from sourceRun the ScienceX CLI from source
Configure AI chatConfigure a model provider
Connect OpenAI, DeepSeek, or a local modelConnect a third-party model
Contribute to ScienceXContributing and quality gates
  1. Download and start the desktop app.
  2. Open Science in the sidebar and run the built-in teaching case. It uses locally generated simulated data and does not require model credentials.
  3. Inspect the research question, input version, run record, and generated artifacts.
  4. Configure a model provider under Settings → Providers only when you want to use AI chat.
  5. Create a project in your own research directory and begin with a task whose scope and expected output are easy to verify.

Local-first does not mean fully offline. Local experiment analysis does not send table contents to a model. When you use a remote model, the prompt, context, and tool results required for the request are sent to the endpoint you configured. Review the provider's data policy and your authorization boundaries first.

Continue learning ​

Released under the MIT License.