AI sandboxes and secure execution tools
Isolated sandboxes where agents can run code and tools away from your machine and data. Of the 137 tools on our agent security and safety page, 9 fit this narrower job. By pricing: 9 open source. The most-starred open repos here are OpenAI Agents SDK (Python) (30,393 stars), NVIDIA OpenShell (20,733 stars), Agent Governance Toolkit (6,495 stars). Most come from the sandboxes & code execution (3), security & governance (3), MCP servers (2) categories.
Updated
OpenAI Agents SDK (Python)
The OpenAI Agents SDK for Python is an MIT-licensed, lightweight framework for multi-agent workflows. Agents are LLMs with instructions, tools, guardrails and handoffs; sandbox agents work inside a container over long time horizons; realtime and voice agents cover speech pipelines; and built-in tracing shows each run. It is provider-agnostic, working with the OpenAI Responses and Chat Completions APIs and 100+ other LLMs, and has a JavaScript/TypeScript sibling (openai-agents-js).
Added
NVIDIA OpenShell
OpenShell is NVIDIA's Apache-2.0 safe, private runtime for fleets of autonomous AI agents: each agent runs in an isolated sandbox where declarative YAML policies control which files, networks, packages, APIs and credentials it can touch. The project is moving to the CNCF (ai-openshell); the 0.1.x line, at v0.1.8 on Nov 10, 2026, brought a stable release cadence, new isolation primitives, an expanded extension surface and new APIs.
Added
Agent Governance Toolkit
Microsoft toolkit for policy enforcement, identity and sandboxing for AI agents.
Added
Redamon
Open-source, self-hosted AI penetration testing framework: maps your attack surface into a graph, autonomously exploits it from a Kali sandbox with human approval gates, and opens PRs that fix what.
Added
Mantis
Google's toolkit for AI coding agents to autonomously find, reproduce and patch software vulnerabilities in isolated environments.
Added
ToolHive
Stacklok platform that runs MCP servers in isolated containers with per-request identity, access policy and observability.
Added
Agent Safehouse
macOS sandbox for local coding agents that limits file and integration access with deny-first policy profiles.
Added
Perplexity SPACE
Perplexity's open-source sandbox platform that runs each agent task in its own Firecracker microVM, with credentials and network access controlled from outside.
Added
Sandlock
The lightest AI sandbox. A process-based sandbox for Linux, no container, no VM, no privilege, no prompt injection
Added
GitHub repos (9)
| Repo | GitHub stars | Licence | Last push |
|---|---|---|---|
| 30,393 | MIT | 2026-10-27 | |
| 20,733 | Apache License 2.0 | 2026-10-30 | |
| 6,495 | MIT | 2026-10-29 | |
| 3,246 | — | 2026-10-28 | |
| 2,675 | Apache 2.0 | 2026-10-25 | |
| 2,316 | Apache 2.0 | 2026-10-30 | |
| 2,128 | Apache 2.0 | 2026-10-19 | |
| 2,007 | Apache 2.0 | 2026-10-30 | |
| 591 | — | 2026-10-28 |
9 repos. Stars and licence come from each repo, checked when the listing was last verified.
FAQ
What are the best AI tools for agent security and safety focused on sandboxes and secure execution?
We list 9. The first three by GitHub stars and featured status are OpenAI Agents SDK (Python): The OpenAI Agents SDK for Python is an MIT-licensed, lightweight framework for multi-agent workflows. Agents are LLMs with instructions, tools, guardrails and handoffs; sandbox agents work inside a container over long time horizons; realtime and voice agents cover speech pipelines; and built-in tracing shows each run. It is provider-agnostic, working with the OpenAI Responses and Chat Completions APIs and 100+ other LLMs, and has a JavaScript/TypeScript sibling (openai-agents-js).; NVIDIA OpenShell: OpenShell is NVIDIA's Apache-2.0 safe, private runtime for fleets of autonomous AI agents: each agent runs in an isolated sandbox where declarative YAML policies control which files, networks, packages, APIs and credentials it can touch. The project is moving to the CNCF (ai-openshell); the 0.1.x line, at v0.1.8 on Nov 10, 2026, brought a stable release cadence, new isolation primitives, an expanded extension surface and new APIs.; Agent Governance Toolkit: Microsoft toolkit for policy enforcement, identity and sandboxing for AI agents.
Are there free or open-source options?
Yes: 9 open source. Pricing is as each maker states it.
More from agent security and safety
Tools are grouped by what each says it does. Missing one? Submit it free.