Sandboxing Python Packages

pydantic-monty

A minimal, secure Python interpreter written in Rust for use by AI

pydantic-monty-runtime

A minimal, secure Python interpreter written in Rust for use by AI

pydantic-monty-client

A minimal, secure Python interpreter written in Rust for use by AI

pydeno

Run AI-generated JavaScript from Python, securely, in a sandbox: real V8 in a supervised, OS-sandboxed worker process (Seatbelt, Landlock, seccomp), with your Python functions as the only way out.

arena-pytest

A cross platform sandboxing framework with built in profiling workflows.

sandlock

The lightest AI sandbox. A process-based sandbox for Linux, no container, no VM, no privilege, no prompt injection

folsom-fuse

Lightweight microVMs for running anything.

locki

Locki: AI sandboxing without the taste of sand

ailoy-py

AI agent builder with a VM at its heart. Every agent gets its own microVM sandbox to install software, run code and use the GPU in. Python, Node.js and Rust.

portmark

Provider-neutral, capability-secured portable agent reference runtime: signed agent envelopes, host-enforced permits, Wasm capsule execution, and checkpoint-based migration.

akios

AKIOS runtime for secure AI agent execution

declaw

Python SDK for Declaw — security-first sandboxing for AI agents

druks

Druks runs agent apps on your own infrastructure: durable runs that survive restarts, human approvals, and full history. Apps install as Python packages.

enforcecore

The runtime enforcement layer for agentic AI systems.

gsj-harness-rollout-server

Rollout server for RL over a temporally-scoped corpus: (case, timestep, prompt) → a pinned agent in a sandbox truncated at that timestep → one validated, training-ready trajectory.

virtx

Disposable Linux VMs for your own code. Runs microVMs on Linux, macOS and Windows hosts with no Docker or daemon, GPU through Vulkan, and host folders mounted in. Python, Node.js and Rust.

torve

A specification-and-gate engine for a standing agent team: graded decisions, scope as a machine-checked contract, deterministic gates, sandboxed agent runs

bandsox

Sanboxes for AI agents and humans

codesafe

An open-source Python library for code encryption, decryption, and safe evaluation using Python's built-in AST module, complete with allowed functions, variables, built-in imports, timeouts, and blocked access to attributes.

ralonlock

Enforce filesystem boundaries for autonomous coding agents — regardless of the tools or commands they use.

mainsheet

An agent runtime with guardrails. Define an agent in one file, run it governed, prove what it did: a policy gate at the tool-call boundary, sandboxing, budgets, OpenTelemetry traces, a signed audit trail, and taxonomy-driven evals. Built on the Claude Agent SDK.

sandwine

:wine_glass: Command-line tool to run Windows apps with Wine and bwrap/bubblewrap isolation on Linux

ectoledger-sdk

A dashcam and emergency brake for AI agents (prototype): a Rust harness that runs agent actions through policy checks and OS sandboxes, and records them in a tamper-evident, signed, independently verifiable audit log.

linura

Linura — The intelligent system layer for Linux, AI agents, containers, and VMs.