Skip to content
DocumentGetting started

Environment & models

2 min read

Prepare your Agent runtime and connect a model for your research.

On this page

Check the environment#

Open Environment in the top-right corner of the WebUI and check that the Agent runtime and model you need are available.

Harness means the Agent program that executes your tasks. When creating a project, choose an available Harness and a compatible model source. Combinations marked unavailable need installation, authentication, or configuration before use.

Choose a model source#

Common model sources include:

  • Codex Subscription: use an authenticated Codex subscription.
  • Local API / Server API: connect a model API you provide. The label depends on where EvoLab runs.
  • Hugging Face: deploy an open model. Check its hardware and runtime requirements, including available GPU memory and disk space when using a GPU model.

Use your own API#

Prepare the provider, API protocol, Base URL, Model ID, and API key. Follow the page or launcher prompts to configure and verify them. When adding an API model in Server mode, you may need to return to the launcher terminal to enter your key without displaying it. Use the exact model name and protocol supported by your provider.

API verification, Research tasks, and Evolution runs may incur model usage charges, according to your provider's billing.