fab.run

What It Is

Fabric is a lightweight sandboxing tool for AI Agent workloads, positioned as “one interface, any runtime.” It provides isolated environments for agents to run code, and connects through a unified interface to local containers, macOS VMs, Daytona, E2B, exe.dev, or a user’s own runtime. For teams that need LLM Agents to execute scripts, debug code, or handle automation tasks, it is more of a runtime adapter layer than a standalone cloud sandbox service.

Core Capabilities and Ecosystem

Judging by its documentation structure, Fabric covers Overview, Getting Started, Philosophy, Architecture, Runtimes, API Reference, and Skills. Its core capability is portable sandboxing for agentic workloads: local containers can be used during development, while cloud environments can connect to Daytona, E2B, and exe.dev, with the option to extend to custom runtimes. The API Reference explicitly mentions TypeScript interfaces, indicating that its SDK is at least aimed at TypeScript developers. It also provides prebuilt Skills for AI agents, as well as llms.txt and llms-full.txt, making it easier for large language models to read the project documentation.

Pricing and Deployment

The captured text does not disclose pricing, free quotas, commercial plans, or payment methods, so the actual cost cannot be assessed. In terms of deployment, the text explicitly lists Local Container, macOS VM, and user-owned runtimes, which suggests it can be used in local or custom environments. It can also integrate with cloud sandboxes such as Daytona, E2B, and exe.dev. However, the text does not clarify whether it supports a fully self-hosted control plane, authentication, auditing, or enterprise deployment.

Pros and Cons

The main advantage is its clear abstraction layer, which reduces the cost of switching between different sandbox backends and makes it suitable for teams building AI Agent toolchains. The documentation entry points also appear fairly complete, including quick start, architecture, and runtime explanations. The downside is that many key details are missing: there is no indication of whether it is open source or closed source, what license it uses, its pricing, support options, or stability commitments. Enterprise-level details such as security boundaries, resource limits, log auditing, and related controls are also not visible.

Who It’s For and Access from China

Fabric is suitable for AI Agent developers, developer tooling teams that need to safely execute untrusted code, and projects that want portability between local and cloud sandboxes. The text does not provide information on access from mainland China. When external services such as E2B, Daytona, and exe.dev are involved, network connectivity and payment availability may also be factors, so real-world testing is recommended. Possible alternatives or complementary options include using E2B, Daytona, or exe.dev directly, or building an execution sandbox with Docker/containers.