Agent developers who need to let an LLM safely execute untrusted, model-generated code.

An open source platform providing secure, isolated microVM sandboxes where AI agents can run code, install packages, and use real tools like a browser or filesystem.

Sandboxes run in E2B's cloud infrastructure, provisioned on demand via SDK calls from any backend.

Use when an agent framework needs a code interpreter tool that isolates execution from the host system.

It lets agents run arbitrary generated code without risking the developer's own infrastructure.

Provides SDKs for Python and JavaScript and integrates with most major agent frameworks.

Open source SDK and sandbox images, Apache 2.0 licensed, with a hosted cloud service requiring an API key for scale.