Gradient Starter Kit — AI skill for Claude Code

DigitalOcean Gradient AI - Quick Reference Guide

If you're looking to get started with [DigitalOcean Gradient](https://docs.digitalocean.com/products/gradient-ai-agentic-cloud/) and want to quickly start building, this is a collection of examples, code snippets, and guides to help you hit the ground running.

Gradient gives you **serverless inference** — one API key that lets you use models from **OpenAI, Anthropic, DeepSeek, Llama, Qwen**, and others — all from a single endpoint.

---

What You Can Do with Gradient

**DigitalOcean Gradient** provides two core AI platform offerings:

1. **Inference** — Run models like GPT-4, Llama 3, DeepSeek, and others directly via a unified API endpoint. Ideal for quick text, image, and embedding tasks.

2. **Agents** — Create persistent, context-aware agents connected to your own data — docs, URLs, buckets, or files. These agents “remember” what you’ve uploaded and respond with that context built in.

---

Three Ways to Use Gradient

You can interact with either **Inference** or **Agents** in any of these ways:

Method	Description	Example Section
cURL / REST API	Use raw HTTP requests — fastest way to test and explore the API.	Section 0 → List Models / Section 1 → Simple Chat
OpenAI SDK	Use the official OpenAI client and simply override the base_url. Fully compatible with the OpenAI API spec.	Section 3 → Using Python (OpenAI SDK)
Gradient SDK	Use DigitalOcean’s native Python client (gradient) — includes both Inference and Agents.	Section 4 → Using Python (Gradient SDK)

All three methods are API-compatible — you can switch between them with minimal code changes.

---

Quick Start Summary

Here’s a quick map of what you can do and where to find it in this guide:

| Goal | Endp