Skip to content

Entry

OpenShell

Appears in 5 awesome lists

Open-source policy-driven sandbox runtime for autonomous AI agents, announced at GTC 2026. Enforces security constraints at the kernel level via Landlock LSM (filesystem), seccomp BPF (syscalls), and an OPA/Rego-evaluated HTTP CONNECT proxy (network) — constraints are enforced on the environment…

Open github.comnvidia/openshell

Found in these lists

Awesome AI Security Tools

Section: Runtime Protection & Enforcement · 🟢 — Policy-governed runtime for autonomous and coding agents with container or microVM-backed sandboxes, filesystem/process/network controls, endpoint-bound credential injection, and audit logs. (NVIDIA) — note: pre-release runtime whose effective isolation depends on the selected compute…

FreshScore 91

Awesome Harness Engineering

Section: Security, Sandbox & Permissions · Open-source policy-driven sandbox runtime for autonomous AI agents, announced at GTC 2026. Enforces security constraints at the kernel level via Landlock LSM (filesystem), seccomp BPF (syscalls), and an OPA/Rego-evaluated HTTP CONNECT proxy (network) — constraints are enforced on the environment…

FreshScore 93

Awesome local LLM

Section: Security and Sandboxing · the safe, private runtime for autonomous AI agents from NVIDIA

FreshScore 90

Awesome Open Source AI

Section: 4. Agentic AI & Multi-Agent Systems · Safe and private runtime for autonomous AI agents with policy-driven execution boundaries and CLI integration.

FreshScore 87

Awesome Prompts

Section: Sandboxed Execution · NVIDIA's security-first runtime for autonomous agents — kernel-level sandboxed execution with policy-based governance over file access, syscalls, and every network connection; credential proxying (agents never see real secrets — injected only at approved endpoints), and formal verification of…

FreshScore 96

Bifrost

🟢🟠 — Apache-licensed AI and MCP gateway with multi-provider routing, virtual-key access controls, budgets, rate limits, MCP aggregation, OAuth, automatic fallbacks, and load balancing. (Maxim) — note: model-provider credentials and proxied request/response data are sensitive; restrict and…;…

In 11 listsDetails

Future AGI

🟢🟠 — Self-hostable evaluation and tracing platform with an OpenAI-compatible Go gateway for inline model-traffic checks, MCP/tool permissions and optional external guardrail services. (Future AGI) Caveat: the core is Apache-2.0; enterprise components have a separate commercial license. Gateway…;…

In 9 listsDetails

E2B

Open-source sandboxed code execution for AI agents — Firecracker microVMs with dedicated kernels per sandbox (~150 ms cold start), fork/snapshot primitives (pause/resume with full memory state), Python/JS SDKs + MCP server, self-hostable via Terraform on AWS/GCP; used by Perplexity, Hugging Face,…

In 9 listsDetails

Portkey AI Gateway

🟢🟠 — Open AI gateway with provider routing, fallback and retry controls, guardrail integrations, observability, and MCP traffic support for model and agent applications. (Portkey) — note: the gateway is general infrastructure rather than a standalone security scanner; model-provider…

In 7 listsDetails

brood-box

🟢 — Experimental runner for coding agents in hardware-isolated microVMs with copy-on-write workspace snapshots, egress profiles, selective secret forwarding, and file-by-file review before applying changes. (Stacklok) — note: APIs and behavior are explicitly experimental; workspace-mode=direct…

In 6 listsDetails

NeMo Guardrails

NVIDIA's programmable guardrails toolkit: define input, dialog, retrieval, execution, and output rails that intercept the agent loop at five distinct layers using the Colang DSL. The execution rail layer specifically governs what tools the LLM can invoke and what their inputs/outputs may contain —…

In 6 listsDetails

mu

Open-source coding agent harness built on pi that moves routine loop decisions into a "judgment kernel": a small, fast judge model answers bounded questions at 40 decision points, covering which tool output enters the context, command risk and permission approvals, prompt injection in web pages…

In 6 listsDetails

microsandbox

Rust/Apache-2.0 local-first microVM runtime for untrusted agent workloads: <100ms boots, OCI-image compatible, fork-and-snapshot of live sandboxes, and an embeddable library API that spawns VMs in-process with no daemon. The distinguishing designs are secret handling — keys never enter the VM, so…

In 5 listsDetails