Introducing Daytona Cloud: The Agent-Native Infrastructure

Last week, we announced Daytona's new strategic focus on providing purpose-built runtimes for AI agents. Today, we're thrilled to introduce Daytona Cloud.

At its foundation, Daytona delivers five key capabilities:

• Fast environment startup times: Sub-90 millisecond sandbox spin-ups, optimized for agent workflows that require rapid iteration.
• Native Docker compatibility: No proprietary formats or custom configurations. Standard container workflows work out of the box.
• Multi-region support: Deploy environments globally with low-latency points-of-presence.
• Stateful sandboxes: Filesystem, environment variables, and process memory are preserved across agent interactions.

Daytona Cloud was not adapted from human-centric cloud models. It was built specifically for a future where intelligent agents interact directly with infrastructure, programmatically and independently.

We didn't just adapt existing infrastructure; we reimagined what compute should look like in an era where intelligence, not humans, will be the primary consumer of these resources.

Goran Draganic, Chief Architect

Performance Benchmarks

Measured benchmarks demonstrate the advantages Daytona Cloud provides over traditional cloud providers:

Performance was measured under typical agentic workloads, focused on fast iteration and short-lived task execution.

Built for Agent Workflows

Daytona Cloud aligns closely with how real-world AI agents behave:

• Agents often create and manage multiple environments concurrently.
• Code is generated and refined progressively, not deployed as a single final artifact.
• Environments need to persist state across steps and iterations.
• File operations, Git actions, shell commands, and code execution must be accessible programmatically.

Every action an agent needs is exposed through a clean, consistent API, allowing full control without human involvement.

Getting Started

Accessing Daytona Cloud is straightforward:

# Python
pip install daytona_sdk

# TypeScript
npm install @daytonaio/sdk

Pricing: Designed for Agent Economics

Daytona Cloud’s pricing model is designed for agent-driven usage patterns:

• Per-second billing with millisecond precision
• No minimum runtime commitment
• Bandwidth costs included, no hidden transfer fees
• A generous free tier available for early adopters

See the full pricing details here.

Infrastructure for an Agent-Native Future

The shift from human-driven software to autonomous agent systems is underway. In a world where trillions of sandbox environments will be created and destroyed dynamically, general-purpose cloud platforms are no longer sufficient.

Daytona provides infrastructure purpose-built for this new model of computing. Fast, stateful, programmatic, and designed from the start for intelligent agents.

Daytona Cloud is available today (sign up or explore the documentation).