Daytona: Building the Agent Cloud for AI Execution
Daytona provides secure, scalable computing environments designed for AI agents. Its platform gives agents isolated sandboxes where they can run code, use Git, interact with desktop applications, access computing resources, and execute multi-step tasks without relying on a shared, uncontrolled runtime.
Visit Daytona ↗Daytona at a Glance: Company Overview
Daytona is an AI infrastructure company focused on the execution layer needed to operate autonomous agents. AI agents increasingly need to do more than generate text. They need to run programs, install packages, inspect files, use browsers, interact with desktop applications, execute tests, and work with external tools. These activities require computing environments that can be created quickly, isolated from other workloads, and managed programmatically.
Daytona addresses this requirement through sandboxes, which function as dedicated computing environments for agents. Depending on the configuration, a sandbox can provide its own filesystem, network stack, allocated CPU and memory, and an operating system suited to the task. Developers can create environments through APIs and software development kits, run workloads inside them, and manage their lifecycle as work progresses.
The company was founded in 2023 by Ivan Burazin, Vedran Jukić, and Goran Draganić. The founding team previously built Codeanywhere, a cloud development environment company. Daytona initially focused on development environment management before shifting toward purpose-built infrastructure for AI-generated code and autonomous agents.
On October 20, 2026, Daytona announced a $60 million Series B led by FirstMark Growth. The round valued the company at $1 billion and brought its total funding to $91 million. Daytona says its platform serves model developers, AI research organizations, and companies deploying agents in real products.
| Company | Daytona |
| Founded | 2023 |
| Industry | AI infrastructure, cloud computing, developer tools |
| Core Product | Secure execution environments and sandboxes for AI agents |
| Co-founders | Ivan Burazin, Vedran Jukić, and Goran Draganić |
| CEO | Ivan Burazin |
| CTO | Vedran Jukić |
| Chief Architect | Goran Draganić |
| Latest Funding | $60 million Series B |
| Total Funding | $91 million |
| Reported Valuation | $1 billion |
| Series B Lead Investor | FirstMark Growth |
| Official Website | daytona.io |
| Developer Documentation | Daytona documentation |
What Does Daytona Do?
1. Provides Computers for AI Agents
Daytona gives agents isolated environments in which they can execute code, install dependencies, work with files, run processes, and use development tools. These environments provide the execution resources an agent needs to carry out tasks beyond generating a response.
2. Makes Agent Execution Scalable
Agents can create and use environments programmatically rather than depending on a developer to prepare each machine manually. Daytona is designed to support parallel workloads, short-lived tasks, persistent sessions, and changing levels of demand.
3. Separates Execution from Agent Reasoning
Daytona focuses on where an agent’s work runs rather than serving as the agent’s reasoning model. Developers can connect their preferred agent frameworks and models to Daytona’s infrastructure and use it as the execution layer beneath their applications.
Daytona Products and Key Capabilities
Isolated Sandboxes
Sandboxes are the core of Daytona’s platform. They provide separated execution environments with their own resources and operating context. Depending on the sandbox type, isolation can be provided through containers or virtual machines.
Fast Environment Startup
Daytona’s documentation describes sub-90-millisecond startup for its sandbox environments. Fast provisioning helps agent applications create execution environments on demand instead of waiting for traditional machines to boot.
Persistent State and Snapshots
Supported environments can preserve files and working state between operations. Snapshotting and lifecycle controls help developers resume work, reproduce environments, and manage tasks that continue across multiple agent interactions.
Computer Use
Daytona provides features for interacting with desktop environments through screenshots and input events. This can support agents that need to operate graphical interfaces or work with software that does not expose a convenient API.
GPU Sandboxes
GPU-enabled environments support workloads such as CUDA development, graphics, and other GPU-accelerated computing. They give developers an option for agent tasks that require more than standard CPU-based execution.
APIs, SDKs, and CLI
Developers can manage environments programmatically through Daytona’s APIs, command-line interface, and SDKs for languages including Python, TypeScript, Ruby, Go, and Java. This makes it possible to integrate sandbox creation and execution into agent applications and developer workflows.
How Daytona’s Agent Cloud Works
Daytona acts as the execution environment for an AI system. An agent decides what task to perform, while Daytona supplies the computing environment in which code, tools, and other operations can run. Developers manage this interaction through APIs, SDKs, or supported integrations.
1. Create an Environment
An application requests a sandbox through the Daytona API or an SDK. The developer can configure the environment, select an appropriate sandbox type, and specify the resources needed for the workload.
2. Prepare the Runtime
The sandbox provides an execution environment with the required operating system, filesystem, packages, and resources. Developers can use compatible images or environment configurations to make execution more consistent.
3. Execute Agent-Generated Work
The agent can run code, execute shell commands, interact with files, use Git, start processes, and work with supported tools inside the sandbox. This separates the execution of generated code from the application that coordinates the agent.
4. Interact with Applications
For tasks involving graphical interfaces, supported environments can expose desktop interaction capabilities. Agents can use screenshots and input events to interact with applications that require mouse, keyboard, or visual navigation.
5. Preserve or Branch Work
Where supported by the sandbox type, developers can pause, resume, snapshot, or fork environments. These capabilities are useful when an agent needs to revisit a previous state or explore several possible solutions in parallel.
6. Scale and Manage Resources
Applications can create environments for many concurrent tasks and manage their lifecycle through Daytona. Stopping or deleting unused environments helps control resource usage, while monitoring and billing tools support cost management.
Enterprise Use Cases for Daytona
AI Coding Agents
Coding agents need to execute generated programs, install packages, inspect repositories, run tests, and evaluate changes. Daytona provides a dedicated environment for these activities, reducing the need for teams to build their own execution infrastructure.
Automated Software Testing
Testing agents can execute test suites, reproduce errors, inspect application behavior, and validate proposed code changes in isolated environments. Teams can use separate sandboxes for parallel test runs or independent branches of a task.
Browser and Desktop Automation
Agents that navigate websites or operate desktop applications need a place to run browser sessions and interact with graphical interfaces. Daytona’s computer-use capabilities support workflows where an API alone is insufficient.
AI Model Evaluation
Research teams can run agent-generated code and evaluation tasks in controlled environments. Parallel execution can help teams compare approaches, evaluate outputs, and repeat experiments under defined runtime configurations.
Reinforcement Learning
Training and research workflows can require repeated interactions with software environments. Daytona’s sandbox infrastructure can support tasks that need reproducible execution, multiple parallel environments, or GPU resources, depending on the workload.
Data and Code Processing
Agents can execute data transformations, run scripts, inspect files, and process datasets within allocated environments. Resource controls and isolation can help teams manage workloads generated dynamically by AI applications.
What Differentiates Daytona?
Daytona’s central proposition is that AI agents need an execution environment designed around their operating patterns. Traditional cloud infrastructure can run agent workloads, but teams may need additional systems to manage fast provisioning, isolated execution, persistent state, parallel task branches, and interactive computer use.
| Capability | Daytona’s Approach | Why It Matters |
|---|---|---|
| Execution Environment | Programmatically managed sandboxes for agents | Allows agents to execute code and use tools in a dedicated environment |
| Startup Speed | Sub-90-millisecond sandbox startup is documented for supported environments | Helps reduce waiting when tasks require environments on demand |
| Isolation | Separate execution environments, with container or virtual-machine options | Helps limit the impact of individual workloads and separate execution contexts |
| Operating Systems | Linux, Windows, macOS, and GPU-enabled environments across supported offerings | Supports a wider range of agent tasks than a single standard runtime |
| State Management | Persistence, snapshots, and lifecycle controls where supported | Enables long-running tasks, reproducible environments, and recovery workflows |
| Parallel Execution | Programmatic creation and management of many sandboxes | Supports simultaneous agent jobs, testing, and evaluation workloads |
| Developer Integration | APIs, SDKs, CLI, and compatible development workflows | Lets teams integrate execution infrastructure with existing agents and applications |
| Resource Management | Configurable compute resources and usage-based billing | Helps teams align infrastructure consumption with actual workloads |
These capabilities describe Daytona’s product positioning and documented features. The appropriate comparison with other infrastructure providers depends on workload type, isolation requirements, operating-system needs, pricing, and performance under real production traffic.
Daytona Pricing and Billing
Daytona’s current documentation describes usage-based billing for sandbox resources. Costs depend on the resources allocated, sandbox type, and lifecycle state. The company’s billing documentation explains how CPU, memory, and disk usage are measured, with details varying by the applicable platform and sandbox configuration.
Compute Resources
CPU and memory are billed according to the applicable resource reservation or usage rules. The exact calculation depends on the selected platform version and sandbox configuration.
Storage and State
Disk usage can contribute to the bill even when a sandbox is stopped or paused, depending on the product and storage configuration. Teams should account for persistent environments and snapshots when estimating costs.
Usage Controls
Daytona provides organization-level spending information and resource breakdowns to help teams track active sandboxes and identify which workloads drive costs.
Daytona’s documentation includes multiple platform versions, and billing behavior can differ between them. Check the current official documentation and billing guidance for applicable rates and resource rules before estimating production costs.
Security and Execution Isolation
Running AI-generated code introduces risks because the code may be incorrect, unpredictable, or potentially harmful. An agent may need to access files, execute commands, install dependencies, or interact with external systems. Daytona’s sandbox model provides a separate runtime in which such operations can be carried out with configured resource and access boundaries.
Isolated Execution
Daytona sandboxes provide separated execution environments. Its virtual-machine sandbox options include dedicated kernels, filesystems, network stacks, and allocated resources. The isolation guarantees depend on the sandbox type and configuration.
Scoped Access Tokens
Daytona documents short-lived access tokens that can be restricted to selected operations. This can help developers provide a task or collaborator with limited access instead of sharing a broader API credential.
Network and Resource Controls
Sandbox configuration and application-level controls can be used to manage resources and access. Production deployments should restrict network destinations, credentials, and permissions according to the workload’s requirements.
Lifecycle Management
Developers can stop, pause, resume, or delete supported environments. Managing lifecycle states helps reduce unnecessary resource use and can limit the period during which a workload remains active.
Controlled Tool Access
Teams should grant agents only the tools and credentials required for a task. Sandboxing is one layer of a secure architecture, not a substitute for authorization checks, secrets management, monitoring, or human approval for sensitive operations.
Enterprise Governance
Organizations should review access policies, data handling, logging, retention, and deployment options before placing sensitive workloads on any execution platform. Daytona’s available configurations and contractual assurances should be verified directly for each deployment.
Daytona Funding History and Investors
Daytona announced a $60 million Series B in October 2026 at a $1 billion valuation. The company said the round brought its total funding to $91 million. Its previous major publicly announced round was a $24 million Series A in February 2026.
Led by FirstMark Capital, with participation from Pace Capital, Upfront Ventures, Darkmode, E2VC, and strategic investors.
Led by FirstMark Growth, with participation from new and existing investors, including strategic investors Qualcomm Ventures, Samsung Next, and Okta Ventures.
Daytona reported $91 million in total funding following its Series B. The $7 million difference between this total and the two announced rounds above reflects earlier funding not detailed in this summary.
The Series B announcement also named individual backers including Patrick Collison of Stripe, Tobi Lütke of Shopify, Aravind Srinivas of Perplexity, Eric Yuan of Zoom, Olivier Pomel of Datadog, Adam D’Angelo of Quora, and Brendan Foody of Mercor.
Founders and Leadership Team
Ivan Burazin
Co-founder and CEO
Burazin previously co-founded Codeanywhere, a cloud development environment platform, and founded the Shift developer conference. He later served as Chief Developer Experience Officer at Infobip. His work has focused on developer tools, infrastructure, and improving software development workflows.
Vedran Jukić
Co-founder and CTO
Jukić co-founded Codeanywhere with Burazin before helping establish Daytona. He leads the company’s technical direction as it builds infrastructure for executing AI-generated code and running agent workloads.
Goran Draganić
Co-founder and Chief Architect
Draganić is a founding member of Daytona and previously worked with the team at Codeanywhere. At Daytona, he serves as Chief Architect, contributing to the design of the platform’s execution infrastructure.
The founders’ earlier experience with cloud development environments informs Daytona’s focus on standardized, programmatically managed computing environments. The company is now applying that experience to a different workload: AI agents that may create, use, and discard computing environments at high frequency.
Customers and Market Traction
Daytona says its customers span model development, AI research, data creation, and applications that use agents in production. The company has named several organizations using its infrastructure, although customer relationships and workloads vary.
Writer
Daytona says Writer has used its platform to provision large numbers of sandboxes. The company reported that Writer provisioned 300,000 sandboxes in a single month, following a period when Writer maintained its own sandbox infrastructure.
Meta and Cohere
Daytona identifies Meta and Cohere among model-building organizations in its customer ecosystem. These types of workloads can require repeatable environments for code execution, model-related workflows, and evaluation tasks.
Postman, Clay, and n8n
Daytona lists Postman, Clay, and n8n among companies building products that use agents. These applications illustrate the demand for execution infrastructure across developer tools, business workflows, and automation.
LangChain and Turing
Daytona’s earlier funding announcement named LangChain and Turing among customers. Such use cases include code execution and other workflows where AI systems need consistent, stateful computing resources.
AfterQuery
Daytona says AfterQuery replaced its in-house Google Kubernetes Engine cluster with Daytona in a day. The company cites this as an example of teams adopting managed execution infrastructure instead of maintaining their own environment systems.
Research and Evaluation Teams
Daytona also serves organizations involved in data creation, model evaluation, and AI research. These workflows can involve many concurrent tasks, repeatable environments, and resource requirements that change with each experiment.
Daytona reported that revenue had grown approximately 25 times since January 2026 and that average month-over-month net revenue retention was approximately 135%, according to its October Series B announcement. These are company-reported metrics, and the company did not disclose the absolute revenue base in that announcement.
Daytona FAQs
1. What is Daytona?
Daytona is an AI infrastructure company that provides secure and scalable computing environments for AI agents. Its platform allows agents to execute code, manage files, use tools, interact with desktop applications, and run other tasks inside programmatically managed sandboxes.
2. What is Daytona’s Agent Cloud?
Daytona describes its platform as an execution layer for AI agents. It provides the computing environments, resources, and lifecycle controls agents need to perform tasks, rather than acting primarily as an AI reasoning model or an agent-management dashboard.
3. How much funding has Daytona raised?
Daytona announced a $60 million Series B on October 20, 2026, bringing its reported total funding to $91 million. The Series B was led by FirstMark Growth and valued the company at $1 billion.
4. Who founded Daytona?
Daytona was founded in 2023 by Ivan Burazin, Vedran Jukić, and Goran Draganić. The founders previously built Codeanywhere, a cloud development environment platform.
5. What is a Daytona sandbox?
A sandbox is an isolated computing environment where an agent can run code and tools. Depending on the sandbox type, it can have its own filesystem, network stack, operating system, CPU, memory, and disk allocation.
6. How quickly can Daytona start a sandbox?
Daytona’s documentation describes sub-90-millisecond startup for supported sandbox environments. Startup performance can vary by configuration and workload, so developers should benchmark the sandbox type and environment they plan to use.
7. Which operating systems does Daytona support?
Daytona documents Linux, Windows, macOS, and GPU-enabled environments across its supported offerings. Availability and technical details vary by sandbox class and platform version.
8. Can Daytona run AI-generated code safely?
Daytona provides isolated environments designed to execute code and agent workloads. Sandboxing can help separate execution contexts, but secure deployment also requires appropriate network restrictions, credential handling, access controls, monitoring, and application-level safeguards.
9. Can Daytona be used for browser and desktop automation?
Yes. Daytona documents computer-use capabilities that allow agents to interact with desktop environments through screenshots and input events. This is useful for tasks that involve graphical applications or websites without suitable APIs.
10. Which programming languages can developers use with Daytona?
Daytona provides SDKs for Python, TypeScript, Ruby, Go, and Java, along with APIs and a command-line interface. Developers can use these tools to create sandboxes, execute code, and manage environments from their applications.
11. How does Daytona charge customers?
Daytona documents usage-based billing for sandbox resources. Charges depend on the platform version, sandbox type, resource allocation, and lifecycle state. Customers should check the current billing documentation for applicable rates and calculations.
12. Which companies use Daytona?
Daytona has named organizations including Writer, Meta, Cohere, Postman, Clay, n8n, LangChain, Turing, and AfterQuery in its public materials. The company’s customer base includes model developers, AI research organizations, and businesses deploying agent-powered applications.
13. What is Daytona’s reported valuation?
Daytona announced a $1 billion valuation alongside its $60 million Series B in October 2026. This is the valuation reported by the company with the funding announcement.
14. How is Daytona different from a traditional cloud provider?
Daytona focuses on execution environments designed around agent workloads, including fast provisioning, isolated sandboxes, programmatic lifecycle management, persistent state, and computer-use features. Traditional cloud services can also run agent workloads, but teams may need to build or integrate these capabilities themselves.