DigitalOcean

What is DigitalOcean?

DigitalOcean is a cloud platform that combines compute, managed databases, storage, and AI inference services under a single account. It serves developers and companies building applications or running AI agents across 20 data centers in 13 regions. All resources, including GPU servers and serverless inference, share one invoice with private networking between resources in the same region.

What are the features of DigitalOcean?

• Serverless model inference

  Calls over 85 open-source and frontier models through an OpenAI-compatible endpoint, billed per token with scale-to-zero pricing.
• Managed agent hosting

  Runs agent frameworks such as LangGraph, CrewAI, or Claude Code in isolated MicroVMs that pause when idle.
• Inference routing (Inference Router)

  Automatically selects a model per request based on defined cost or speed criteria without requiring code changes.
• GPU compute instances

  Provides Droplets powered by NVIDIA H100, H200, B300, and AMD Instinct MI300X, MI325X, MI350X, and MI355X hardware.
• Custom model import

  Imports Safetensors weights from Hugging Face for use with Dedicated Inference on supported architectures.
• Agent observability tools

  Records agent actions, flags performance shortfalls, and tests changes through Insights, Signals, and Evaluations services.

What are the use cases of DigitalOcean?

• Prototyping AI applications

  Start with Serverless Inference to call foundation models via API and pay per token before committing to dedicated infrastructure.
• Running always-on production agents

  Deploy an existing harness like Codex CLI or Hermes in a Managed Agents MicroVM alongside governed access to over 16,000 tools.
• Hosting large open-weight models

  Run models such as DeepSeek V4 Pro using Dedicated Inference or GPU Droplets while keeping databases and object storage in the same account.

DigitalOcean Pricing Information:

• Serverless Inference

  Pay-per-token billing with scale-to-zero pricing; supports 85+ models.
• Managed Agents

  Compute charges pause when the agent's MicroVM is idle.
• Dedicated Inference and GPU Droplets

  Pricing depends on workload; see current pricing for details.