EvoLab lets you run a research Agent on your own computer or a remote server and reuse what you learn in later work. This guide takes you from installation to your first Evolution run.
Your first workflow#
- Set up your environment. Download and activate EvoLab, choose your computer or a server, and connect an available model.
- Complete a research task. Create a Project, add your materials, and start your first Session in Research.
- Bring experience into your next task. Review and apply improvements in Evolution, then evaluate them with a new research task.
A few concepts to know#
| Concept | Role in your workflow |
|---|---|
| Project | Organizes files, sessions, and improvement records around a research goal. |
| Workspace | Holds the data, code, and documents the Agent uses and produces. |
| Session | A research conversation focused on a specific objective, with room for follow-up instructions. |
| Research | Where you run tasks, follow their progress, and review results. |
| Evolution | Generates reusable instruction and skill improvements from completed tasks. |
| Memory | Organizes task experience in the background for new, independent sessions. |
Choose where to run#
Local runs the Agent and workspace on your computer. Server connects to your own Linux server through SSH. Files and task execution stay on the server while you work in a browser on your computer.
Both modes use the same Research and Evolution workflow. Choose based on where your data lives and the computing resources you need. See First launch for the steps.
Before you begin#
Prepare the invitation key provided by the EvoLab team and decide how you will access a model. If you plan to use a remote server, have its SSH connection details ready.
Your invitation key grants an EvoLab installation license. Prepare the API key or subscription login required to access your chosen model separately.