Hub/AI Security/Sandboxing
Which AI tools have sandboxing?
OS-level or container isolation for agent commands
3 Yes3 Partial
Verified Yes: Claude Code, Codex, Cursor.
| Tool | Status | Notes | Source |
|---|---|---|---|
| Claude CodeAnthropic | Yes | OS-level sandbox for Bash: Seatbelt on macOS, bubblewrap on Linux/WSL2 | Source |
| CodexOpenAI | Yes | OS-enforced sandbox (read-only / workspace-write); network off by default | Source |
| CursorAnysphere | Yes | Shell commands run in a sandbox when possible (Auto-review / Allowlist) | Source |
| Gemini CLIGoogle | Partial | Opt-in sandbox via -s or GEMINI_SANDBOX | Source |
| GitHub Copilot (VS Code)GitHub / Microsoft | Partial | Docs reference agent sandboxing and workspace trust | Source |
| KiroAWS | Partial | Docs list a Local command sandbox feature; details not reviewed | Source |
| ChatGPT (apps & connectors)OpenAI | Unknown | Not found in official docs checked 2026-11-04 | — |
| Claude apps (claude.ai / Desktop)Anthropic | Unknown | Not found in official docs checked 2026-11-04 | — |
| ClineCline | Unknown | Not found in official docs checked 2026-11-04 | — |
| JetBrains AIJetBrains | Unknown | Not found in official docs checked 2026-11-04 | — |
| Windsurf / Devin DesktopCognition | Unknown | Not found in official docs checked 2026-11-04 | — |
| ZedZed Industries | Unknown | Not found in official docs checked 2026-11-04 | — |
Other AI Security features
Weekly, free
What changed in AI integrations
One short email when MCP clients, plugin ecosystems, IDE integrations, AI plan prices or security advisories change. Every item is sourced.