Heads up This site is currently under heavy development.
← all tools
◆ AI Agent Frameworks

NemoClaw

open-source
source ↗ rss ↗

Summary

NemoClaw is an open-source reference stack, free under Apache 2.0, for running AI agents like OpenClaw, Hermes, and LangChain Deep Agents Code more safely inside NVIDIA OpenShell sandboxes. It runs as a CLI installed locally, with an onboarding wizard that handles sandbox setup, managed inference routing, network policy, and lifecycle operations, and it can also be driven through a starter prompt pasted into a coding agent such as Cursor, Claude Code, Codex, or Copilot. It is built for developers deploying always-on agents who need sandboxing and credential handling done for them rather than assembled from scratch, and it sits in the category of infrastructure for running agents safely rather than the agents themselves. First committed in 2026, it already has 196 contributors and over six thousand commits in the past year, putting it firmly in active development.

What NemoClaw answers

Which agents can I actually run through this?

OpenClaw, Hermes, or LangChain Deep Agents Code, chosen during onboarding or interactively at install time

What does it need on my machine to run?

a supported host such as DGX or Windows Subsystem for Linux, plus the OpenShell sandbox runtime it installs alongside itself

Do I have to use the terminal myself?

no, a coding agent like Cursor or Claude Code can run the install and setup commands on your behalf after you approve each one

How does it handle credentials?

a local credential helper collects and stores them outside the chat, and commands only run after you approve the exact one that will receive them

What happens if I don't want the guided setup?

you can decline the express preset and choose the sandbox name, inference provider, and model interactively instead

Is this the agent itself or something around it?

it's the sandboxing, credential, and lifecycle layer that runs an existing agent, not an agent on its own

all 4 features, with the evidence for each →

Features

4 capabilities · 4 backed by code, an API document or a real run

Built from everything we hold on NemoClaw — every release we have summarised, its product documentation and how that documentation has changed, its README, its command-line surface and API, and runs we performed ourselves. Dates are when we first saw a capability, not when the vendor introduced it.

Capability area
All capabilities 4 capabilities
Request and shutdown timeout controls verified Allows users to configure how long the tool waits for individual requests and for graceful shutdown before aborting. 2 other sources · first seen Aug 2026

command line

  • --request-timeout-seconds — request timeout seconds v0.0.106 · Aug 2026 · command-line history
  • --shutdown-timeout-seconds — graceful shutdown timeout seconds v0.0.106 · Aug 2026 · command-line history
Upstream llama-server proxy verified Connects to a running llama-server instance via configurable host and port settings. 2 other sources · first seen Aug 2026

command line

Recovery preparation verified Provides a CLI subcommand to prepare a recovery process. 1 other source · first seen Aug 2026

command line

Replacement scheduling verified Accepts a configurable start time to schedule when a replacement operation begins. 1 other source · first seen Aug 2026

command line

Capability
Evidence

Lines in monospace are the tool's own words — help text parsed from its source, or an endpoint from its API document. Everything else is our summary of a dated release or documentation change, linked back to the source it came from.

Release history

Nothing shipped since we started watching. Releases and docs updates land here as they happen.

watching since Jul 12, 2026

my-toolchain — 0 tools
paste an install list to detect your tools

A brew list, a Brewfile, requirements.txt, a Dockerfile — or just the product names, free-form. Nothing leaves your browser.

    browse all tools →