#2121st of 21 in Sandboxes
code-interpreter
Runs AI-generated code in isolated cloud sandboxes via JS and Python SDKs
- Stars
- 2.4k
- License
- Apache-2.0
- Last commit
- Nov 2026
- Last release
- Aug 2026
- Language
- Python
2,279 on 25 Apr 20262,430 on 14 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
E2B is hiring: 15 open roles
Overview
E2B Code Interpreter lets an app start isolated cloud sandboxes and run AI-generated code in them through JavaScript/TypeScript and Python SDKs. Variables persist across runCode calls within a sandbox. This repository now holds only the sandbox template, including a chart data extractor; the SDK sources moved to the E2B monorepo.
Who it is for: Developers giving LLM agents a safe place to execute code
Strengths
- Variables persist between run_code calls in the same sandbox
- Official SDKs for both JavaScript/TypeScript and Python
- Custom sandbox templates can add packages or a different runtime
- Cookbook with examples for several LLMs and agent frameworks
Weaknesses
- Quickstart needs an E2B account and API key; sandboxes run in E2B's cloud
- SDK source and issues live in the E2B monorepo, not this repo
- No Docker or compose setup for self-hosting in this repo
- Sandbox resource limits and pricing are not in the README
What it needs
- no GPU
- Needs E2B account and API key (E2B_API_KEY)
What changed
“8d51873: Bump pillow to 12.3.0 to address multiple security vulnerabilities (heap out-of-bounds writes, decompression bomb bypasses, and DoS issues fixed in 12.3.0)”
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Sandboxes
See all 21| Rank | Project | Score |
|---|---|---|
| 1 | LightpandaHeadless browser written in Zig for agents and scraping, with CDP and MCP | 94 out of 100 |
| 2 | OpenShellPolicy-enforced sandbox runtime for autonomous AI agents | 90 out of 100 |
| 3 | NemoClawReference stack for running OpenClaw, Hermes and Deep Agents in OpenShell sandboxes | 88 out of 100 |
| #4 | opensandboxSandbox API for running AI agent code, commands and browsers in isolation | 88 out of 100 |
| #5 | ObscuraHeadless browser engine in Rust with V8 and a Chrome DevTools Protocol server | 87 out of 100 |