nono
nono is an open source, kernel-enforced sandbox for AI coding agents and LLM workloads. Unlike policy-based guardrails that intercept and filter operations, nono uses OS security primitives — Landlock on Linux and Seatbelt on macOS — to make unauthorised operations structurally impossible at the syscall level.
Wrap any AI agent — Claude Code, OpenCode, OpenClaw, or any CLI process — with a single command. nono applies default-deny filesystem access, blocks destructive commands (rm, dd, chmod, sudo), isolates credentials and API keys, and cascades all restrictions to child processes. No escape mechanism exists once restrictions are applied.
Built-in profiles get you running in seconds. Secrets inject securely from the system keystore and are zeroised on exit. Audit logging, atomic rollbacks, and Sigstore-attested policy signing are on the roadmap.
Apache 2.0. From the creator of Sigstore.
Learn more
OpenFang
OpenFang is an open source Agent Operating System built in Rust that provides a unified runtime for building, deploying, and managing autonomous AI agents at production scale. It packages a batteries-included architecture into a single binary, enabling developers to run agents that operate continuously, build knowledge graphs, and report results to a centralized dashboard without constant user prompts. At the core of OpenFang are “Hands,” pre-built autonomous capability packages that execute on schedules and perform tasks such as lead generation, research, browser automation, and social management. It includes dozens of pre-built agents, native tools, and channel adapters that allow agents to function across platforms like Slack, WhatsApp, Discord, and Teams from a single environment. Security is built into the foundation through multiple defense layers such as WASM sandboxing, cryptographic signing, taint tracking, and tamper-evident audit trails.
Learn more
NullClaw
NullClaw is an ultra-lightweight autonomous AI assistant infrastructure built in Zig and distributed as a single static binary designed to run efficiently on virtually any hardware. It emphasizes extreme performance and minimal resource usage, shipping as a roughly 678 KB executable that typically consumes about 1 MB of RAM and boots in under two milliseconds. It eliminates traditional runtime overhead by avoiding virtual machines, interpreters, and complex dependency chains, allowing developers to deploy agents simply by running the compiled binary. Despite its small footprint, the framework provides a full autonomous agent stack with support for more than 22 model providers, 18 communication channels, hybrid vector and FTS5 memory, streaming, voice, and multi-layer sandboxing. Security is built in through workspace scoping, explicit command allowlists, encrypted secrets, and strict sandbox isolation using tools such as Landlock, Firejail, or Docker.
Learn more
Sprites
Sprites by Fly.io is a stateful sandbox platform for running arbitrary code in hardware-isolated Linux environments. A Sprite acts like a persistent Linux computer where developers can execute AI agents, user-uploaded binaries, scripts, applications, and other workloads. The platform supports checkpoint and restore, allowing environments to persist and resume instead of starting from scratch each time. Sprites use fast directly attached NVMe storage that continuously syncs to durable external object storage. Developers can create, manage, execute commands, and connect to Sprite consoles through the CLI, REST API, JavaScript, Go, Elixir, and Python tooling. Built for developers and AI infrastructure teams, Sprites gives applications a simple place to safely run code with persistence, sandboxing, usage-based pricing, and scalable execution.
Learn more
