E2B Sandbox API

aid: e2b:sandbox-api
name: E2B Sandbox API
description: The E2B Sandbox API creates and controls per-agent Linux microVMs that can run arbitrary
  code, install packages, read and write files, stream terminal output, and host headless browsers. Code
  Interpreter is a higher-level surface tuned for LLM-driven Python and JavaScript execution with rich
  outputs.
humanURL: https://e2b.dev/docs
baseURL: https://api.e2b.dev
tags:
- Sandboxes
- Code Interpreter
- Templates
- Filesystem
- Process
- Terminal
- Desktop
properties:
- type: Documentation
  url: https://e2b.dev/docs
- type: APIReference
  url: https://e2b.dev/docs/sdk-reference
- type: GettingStarted
  url: https://e2b.dev/docs/quickstart
- type: SignUp
  url: https://e2b.dev/auth/sign-up
- type: SDK
  url: https://github.com/e2b-dev/e2b
- type: SDK
  url: https://github.com/e2b-dev/code-interpreter
- type: SDK
  url: https://github.com/e2b-dev/desktop
- type: GitHubOrganization
  url: https://github.com/e2b-dev
- type: GitHubRepository
  url: https://github.com/e2b-dev/e2b-cookbook
- type: Pricing
  url: https://e2b.dev/pricing
- type: Blog
  url: https://e2b.dev/blog
features:
- name: Fast Sandbox Boot
  description: Firecracker microVMs that start in under 200ms with no cold starts.
- name: Multi-Language Execution
  description: Run Python, JavaScript, Ruby, C++, and more inside the sandbox.
- name: Code Interpreter
  description: Higher-level SDK that returns charts, tables, and stdout/stderr for LLM-driven analysis.
- name: Filesystem and Process
  description: Full file I/O, package installation, and process control with streaming output.
- name: Custom Templates
  description: Build reusable sandbox images with your own dependencies and tools.
- name: Long-Running Sessions
  description: Sessions up to 24 hours on Pro for deep research and long agent runs.
- name: Desktop Sandbox
  description: Virtual desktop environments for LLM computer-use agents.
- name: Browser Inside Sandbox
  description: Headless browser access for agents that need both code and web.
- name: Cookbook and Templates
  description: 20+ integration patterns published in the e2b-cookbook repo.
useCases:
- name: LLM Code Interpreter
  description: Power Python/JS code execution behind chat assistants and copilots.
- name: Deep Research Agents
  description: Give agents a persistent, secure workspace for multi-step research jobs.
- name: Data Analysis and Visualization
  description: Let agents load datasets, run analysis, and return charts safely.
- name: Computer Use Agents
  description: Pair with vision models to operate a full desktop environment.
- name: CI/CD and GitHub Actions
  description: Run untrusted or AI-generated code inside isolated VMs.
integrations:
- name: OpenAI
- name: Anthropic
- name: Mistral
- name: Llama
- name: LangChain
- name: LlamaIndex
- name: Vercel AI SDK
- name: CrewAI
- name: AutoGen
- name: Hugging Face
authentication:
- type: API Key
  description: API key passed via the `E2B_API_KEY` environment variable or `Authorization` header authenticates
    SDK and REST calls.