Getting Started

Install, authenticate, and create your first autonomous DeepBox workflow.

1. Install the CLI

Run `npm install -g deepbox` or use the Docker image to get the latest tooling for orchestrating experiments.

2. Authenticate

Generate an API token from the DeepBox dashboard and run `deepbox login --token <TOKEN>` to authorize the CLI.

3. Launch your first agent

Use `deepbox init` to scaffold a project, point it at your dataset, and run `deepbox up` to deploy an autonomous research loop.

4. Explore project scaffolding

The generated project includes example configs, workflow templates, and optional GitHub Actions to automate retraining. Tweak the YAML files and rerun `deepbox up` to iterate quickly.

5. Visualize progress

The CLI streams structured logs and optionally uploads to DeepBox Studio, where you can view traces, compare checkpoints, and attach notes for future experiments.

6. Invite collaborators

Run `deepbox team add <email>` to share access to agents, datasets, and experiment history. Role-based permissions ensure production safeguards remain intact.

7. Roll back safely

Use `deepbox rollback <deployment>` to revert to a previous checkpoint while keeping the current workflow configuration intact for future tweaks.

8. Customize templates

DeepBox includes curated workflow templates. Clone the repo, tweak the YAML, and point your CLI to the local directory to start from best practices.

Was this page helpful?