
Cloudflare's virtual filesystem for AI agents — a Durable Object-backed workspace with three pluggable execution runtimes.
Cloudflare's virtual filesystem for AI agents — a Durable Object-backed workspace with three pluggable execution runtimes.
Cloudflare Computer gives AI agents a computer to run on. The public @cloudflare/computer package exposes a Workspace whose authoritative state lives in a Durable Object's SQLite store, with one pluggable execution surface — workspace.runtime.exec — that lets an agent run shell commands or ECMAScript modules against that filesystem. Three backends ship today: a Container backend that FUSE-mounts the workspace into a real Linux sandbox with full userland and network, an Isolate Shell backend that runs just-bash inside a Dynamic Worker with no second store, and an Isolate JavaScript backend that evaluates ES modules in a fresh Dynamic Worker with structured I/O, durable relative imports, and trusted ws:git / ws:artifacts modules. Multiple backends can register under stable IDs on the same Workspace, and callers can even skip the runtime entirely and just get the filesystem. The project is a preview release with an evolving spec under docs/ and worked examples for container, worker-shell, worker-javascript, egress policies, and a full 'think' agent tutorial.
Cloudflare Computer is a virtual filesystem designed for AI agents, utilizing Durable Object technology to create a workspace that supports three pluggable execution runtimes. This environment allows developers to efficiently run AI applications with improved scalability and performance.
Cloudflare Computer leverages its proprietary Durable Object technology, which allows for persistent storage that retains state and data across different sessions. This is crucial for AI agents that require consistent and reliable access to data. The system's architecture is designed to manage workloads efficiently, making it ideal for applications that need to scale rapidly.
The three pluggable execution runtimes within Cloudflare Computer provide developers with the ability to choose the environment that best suits their needs. This flexibility enables the integration of various programming languages and frameworks, accommodating a range of AI applications, from natural language processing to image recognition.
For example, a developer working on a machine learning model can use one runtime to train the model while simultaneously deploying it on another runtime, ensuring that updates and modifications can be made without downtime. This versatility not only enhances productivity but also enables rapid iteration and deployment of AI solutions.
Cloudflare Computer operates by integrating various components, including Durable Object Workspace for persistent state management, pluggable runtime backends for flexibility, and multiple execution environments like container and isolate backends. This architecture enables efficient coding, testing, and artifact generation within Cloudflare's edge network.
Cloudflare Computer is designed to enhance the development experience on its edge network. Here’s how each component contributes:
Durable Object Workspace: This is at the heart of Cloudflare Computer, where the authoritative filesystem state is stored using SQLite. This ensures that the state is transactional, colocated, and resilient against worker restarts. For example, if a worker crashes, the Durable Object retains its state, allowing for seamless recovery and continuity.
Pluggable Runtime Backends: The architecture allows a single workspace to register multiple execution backends under stable IDs. This means developers can select the most suitable backend for each call. For instance, a workspace can execute tasks in both a container backend, which offers a full Linux environment, and an isolate backend, which is more lightweight and focused.
Container and Isolate Backends:
These components empower developers to create coding agents that can write, test, and deploy code directly within Cloudflare's ecosystem. For example, a think-style chat agent can utilize the Workspace as a scratch directory, ensuring that files created are durable and accessible across different sessions.
Cloudflare Computer offers several key features, including Durable Object Workspace for transactional file management, pluggable runtime backends for flexible execution, and specialized backends like Container and Isolate Shell for enhanced processing capabilities. These features enable developers to build scalable, efficient applications with minimal overhead.
Cloudflare Computer is designed to optimize application performance and scalability through its unique features:
Durable Object Workspace: This feature allows developers to maintain authoritative filesystem states in a Durable Object's SQLite store. It ensures that the agent states are transactional, meaning that all changes are recorded reliably. This state management survives worker restarts, providing resilience in application performance.
Pluggable Runtime Backends: One of the standout features is the ability to register multiple execution backends using a single workspace.runtime.exec entry point. This flexibility allows developers to choose the most suitable backend for each call, enhancing the efficiency of resource usage and execution speed.
Container Backend: This backend projects the SQLite state into a sandbox container, simulating a full Linux userland. It allows developers to run real binaries and access real network functionalities, making it ideal for applications requiring full system capabilities.
Isolate Shell Backend: This backend runs a bash shell in a Dynamic Worker, connecting to the authoritative Workspace via Workers RPC. It eliminates the need for a separate container or sync round trips, streamlining processes for lightweight tasks.
Isolate JavaScript Backend: This backend evaluates ES modules in a Dynamic Worker environment, supporting structured input and results. It also includes durable relative imports and is equipped with trusted modules like ws:git and ws:artifacts, enabling secure and efficient module handling.
Cloudflare Computer is designed for developers and teams looking to create coding agents, compare execution runtimes, run untrusted scripts, and prototype multi-tenant applications. It provides a versatile environment for building and testing applications at the edge, catering to various use cases in real-time software development.
Cloudflare Computer offers a robust platform for developers and teams engaged in a range of software development tasks:
Building Coding Agents: With Cloudflare Computer, developers can create Workers-native agents that operate in a unique environment with their own filesystem and shell. This enables agents to write code, conduct tests, and generate artifacts without leaving the Cloudflare edge. For instance, a developer can build a coding agent that automatically implements updates or runs diagnostics on deployed applications, ensuring seamless integration with existing workflows.
Comparing Runtimes Side-by-Side: The platform features an intuitive compare-runtimes UI, allowing users to run the same task on both container and worker runtimes. This side-by-side comparison helps profile key performance metrics such as latency, isolation, and cost. Teams can leverage this information to make informed decisions about optimizing their applications for better performance and reduced expenses.
Sandboxed User Scripts: Cloudflare Computer provides a secure environment for executing untrusted user-supplied JavaScript. The Isolate JavaScript backend ensures that scripts run in a per-user Workspace filesystem, preventing any harmful interactions with the broader system. This is particularly useful for platforms that allow user-generated content, as it mitigates security risks.
Agent Working Directories: The think-style chat agent can utilize the Workspace as a persistent scratch directory. Files written by the agent are durable across sessions and can be accessed by other backends, facilitating better state management and continuity in conversations or tasks.
Prototyping Multi-tenant Runtimes: Cloudflare Computer enables teams to design agent-runtime products using Durable Objects without committing to a specific backend architecture. This flexibility is crucial for teams exploring innovative solutions in multi-tenant environments, allowing for rapid iteration and testing.
Cloudflare Computer is free to use for developers and businesses, providing a serverless environment to run applications without incurring costs. This makes it an attractive option for startups and small projects that require scalability without the financial burden.
Cloudflare Computer operates on a serverless architecture, allowing developers to deploy applications without worrying about server management. This service is part of Cloudflare's broader suite of tools designed to enhance performance and security for web applications.
Cloudflare Computer can be integrated with other Cloudflare services such as Workers and Pages, enabling developers to create complex applications with minimal overhead.
For more detailed insights, consider exploring Cloudflare's official resources or community forums for tips and shared experiences from other users.
To get started with Cloudflare Computer, visit GitHub Cloudflare Computer. There, you can sign up for an account, access the documentation, and explore examples to help you understand how to leverage Cloudflare's powerful computing platform effectively.
Cloudflare Computer is a serverless computing platform that allows developers to run code at the edge of the network, providing low-latency access for global users. To get started, follow these steps:
Sign Up: Go to GitHub Cloudflare Computer and create an account. You will need a GitHub account to proceed.
Explore Documentation: Familiarize yourself with the official documentation available on the GitHub repository. This includes setup guides, API references, and examples that demonstrate various functionalities.
Set Up Your Environment: Install the required tools, such as Node.js or Python, depending on your preferred programming language. Ensure you have the Cloudflare CLI installed for easier deployment and management of your projects.
Create Your First Project: Use the examples in the documentation to create a simple application. You can start with basic functions like HTTP responses or more complex tasks like interacting with databases.
Deploy and Test: Once your application is ready, deploy it to the Cloudflare network using the CLI. Test your application to ensure it runs smoothly and meets performance expectations.
Browse by use case: Code Generation
Compare Cloudflare Computer: vs Execlave · vs Trigger.dev · vs Cerenovus · vs Portfolio Lab