ClaudeMap

E2B

Open-source secure sandboxes where AI agents run code — the de facto execution layer for agent apps.

SDKs & Toolingsandboxcode-executionsdkagents
Visit projecte2b-dev/E2B

E2B provides isolated sandboxes (fully open source, self-hostable) purpose-built for AI-generated code: spin up a sandbox, let the agent run Python/JS/shell inside it, stream results back. The default execution layer for agent demos and production apps alike, and a natural pairing with the Claude Agent SDK when your agent needs to run, not just write, code.

What it does

  • Isolated sandboxes with fast start, scriptable from Python or JS SDKs
  • Filesystem, network, and process access inside the sandbox — agent code runs for real
  • Self-hostable (fully open source) or used as a managed cloud
  • Optional desktop and code-interpreter sandbox flavors for GUI and data tasks

When to use it

Best whenever an agent must execute untrusted generated code: data analysis, chart rendering, web scraping jobs, CI-style validation of AI patches.

Getting started

pip install e2b             # Python SDK
npm i @e2b/code-interpreter # JS code-interpreter SDK

# or the desktop flavor:
pip install e2b-desktop

Things to know

Sandboxes are ephemeral by design — persist state to object storage between runs. Managed cloud is metered; heavy agent loops with long-lived sandboxes are where costs creep.

Repository

e2b-dev/E2B

Charted

Related on the map

  • FastMCP

    SDK / Tool

    Pythonic framework for building MCP servers — decorators, type hints, and zero boilerplate.

    • framework
    • python
    • mcp
    • type-driven
    View

  • Claude Agent SDK

    SDK / Tool

    Anthropic's official TypeScript SDK for building agentic applications.

    • official
    • sdk
    • typescript
    • anthropic
    View