> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/llms.txt. # NemoClaw for Deep Agents ## Docs - [NVIDIA NemoClaw](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/home.md): NemoClaw is an open source reference stack for running supported and qualified candidate AI agents more safely with guided onboarding, managed inference, policy, and lifecycle operations. - [Overview of NVIDIA NemoClaw](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/about/overview.md): NemoClaw is an open-source reference stack for running sandboxed AI agents more safely inside OpenShell. - [NemoClaw Architecture Overview](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/about/how-it-works.md): Learn how NemoClaw combines a host CLI, agent integration layer, versioned blueprint, and OpenShell gateway to operate supported agents. - [Ecosystem](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/about/ecosystem.md): How LangChain Deep Agents Code, OpenShell, and NemoClaw form one managed terminal-agent stack. - [Release Notes](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/release-notes.md) - [Release Notes (all entries)](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/release-notes/llms.txt) - [Prerequisites](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/get-started/prerequisites.md): Hardware, software, and supported platforms for running NemoClaw. - [Prepare DGX Station to Install NemoClaw](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/get-started/additional-setup/dgx-station-preparation.md): Qualify and prepare a DGX Station GB300 for NemoClaw express install. - [Prepare a Windows Machine to Install NemoClaw](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/get-started/additional-setup/windows-preparation.md): Prepare a Windows machine for NemoClaw before running the Quickstart: enable WSL 2, install Ubuntu, and choose Docker Desktop or qualified rootless Podman for WSL-local Ollama. - [Quickstart with LangChain Deep Agents Code](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/get-started/quickstart.md): Install NemoClaw, launch a LangChain Deep Agents Code sandbox, and run your first prompt. - [About Inference Routing](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/about-inference-routing.md): Understand how NemoClaw routes sandbox inference through OpenShell without exposing provider credentials. - [Choose an Inference Provider](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/learn-and-choose/choose-inference-provider.md): Compare hosted, local, and compatible inference providers available during NemoClaw onboarding. - [Choose a Model](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/learn-and-choose/choose-model.md): Compare curated cloud models by task fit, latency, tool use, context, and relative cost. - [Use NVIDIA Endpoints](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/hosted-inference/use-nvidia-endpoints.md): Configure NemoClaw to use models hosted on NVIDIA Endpoints. - [Use OpenRouter](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/hosted-inference/use-openrouter.md): Configure NemoClaw to route hosted inference through OpenRouter. - [Use OpenAI](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/hosted-inference/use-openai.md): Configure NemoClaw to use the OpenAI API for hosted inference. - [Use Anthropic](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/hosted-inference/use-anthropic.md): Configure NemoClaw to use the Anthropic Messages API for hosted inference. - [Use Google Gemini](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/hosted-inference/use-google-gemini.md): Configure NemoClaw to use Google Gemini through its OpenAI-compatible endpoint. - [Retired Model Router](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/hosted-inference/set-up-model-router.md): Choose a supported provider after Model Router removal and clean up an existing installation. - [Choose a Local Inference Server](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/local-inference/choose-local-inference-server.md): Compare operator-run servers and managed runtime profiles before choosing local inference for NemoClaw. - [Set Up vLLM](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/local-inference/set-up-vllm.md): Connect NemoClaw to an existing vLLM server or start a managed vLLM container. - [Set Up vLLM on Two DGX Sparks](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/local-inference/set-up-vllm-on-two-dgx-sparks.md): Qualify two trusted DGX Spark systems and start NemoClaw's Experimental distributed DeepSeek V4 Flash vLLM profile. - [Set Up vLLM on Two DGX Stations](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/local-inference/set-up-vllm-on-two-dgx-stations.md): Qualify a trusted two-Station DGX GB300 pair and use NemoClaw's distributed Nemotron Ultra vLLM recipe. - [Set Up llama.cpp](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/local-inference/set-up-llama-cpp.md): Attach an existing llama.cpp server or select an experimental managed llama.cpp profile. - [Set Up NVIDIA NIM](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/local-inference/set-up-nvidia-nim.md): Pull, start, and configure a local NVIDIA NIM container for NemoClaw inference. - [Set Up an OpenAI-Compatible Endpoint](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/custom-endpoints/set-up-openai-compatible-endpoint.md): Connect NemoClaw to a self-hosted or custom OpenAI-compatible inference endpoint. - [Set Up an Anthropic-Compatible Endpoint](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/custom-endpoints/set-up-anthropic-compatible-endpoint.md): Connect NemoClaw to a custom Anthropic-compatible inference endpoint. - [Choose a Compatible Inference API](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/custom-endpoints/choose-compatible-inference-api.md): Choose the Chat Completions or Responses API for a custom OpenAI-compatible endpoint. - [Meet Custom Endpoint Security Requirements](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/custom-endpoints/custom-endpoint-security.md): Understand credential isolation and URL validation requirements for custom NemoClaw inference endpoints. - [Use Shared Gateway Routes](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/manage-inference/use-shared-gateway-routes.md): Understand how multiple NemoClaw sandboxes safely share one live OpenShell inference route. - [View the Selected Inference Path](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/manage-inference/view-active-inference-route.md): Inspect a sandbox-attached native provider or the live NemoClaw-managed OpenShell inference route. - [Switch Inference Models](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/manage-inference/switch-models.md): Change the model on a NemoClaw-managed inference route. - [Switch Inference Providers](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/manage-inference/switch-providers.md): Move a NemoClaw-managed sandbox to another inference provider. - [Configure Model Limits](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/manage-inference/configure-model-limits.md): Context-window and output-token limits for a NemoClaw-managed sandbox. - [Configure Inference Timeouts](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/manage-inference/configure-inference-timeouts.md): Configure NemoClaw inference request, provider-validation, sandbox-readiness, and recovery timeouts. - [Understand Provider Validation](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/validate-inference/understand-provider-validation.md): Understand how NemoClaw validates inference credentials, models, APIs, and streaming behavior during onboarding. - [Verify the Sandbox Inference Route](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/validate-inference/verify-inference-route.md): Verify that a NemoClaw sandbox can reach its configured model through its selected inference provider. - [Model Capability Audit Matrix](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/inference/validate-inference/audit-model-capabilities.md): Maintained matrix template for auditing NemoClaw model and provider behavior across supported agent surfaces. - [Configure Progressive Tool Disclosure](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/configure-agents/progressive-tool-disclosure.md): Understand and configure how NemoClaw agents discover tools without loading every schema into model context. - [View Sandbox Status](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/operate-sandboxes/view-sandbox-status.md): List NemoClaw sandboxes, check their health, inspect logs, and collect diagnostics. - [Run Sandboxes](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/operate-sandboxes/run-sandboxes.md): Run multiple sandboxes and stop or start them through OpenShell, with dashboard and tunnel guidance where supported. - [Run LangChain Deep Agents Code](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/operate-sandboxes/run-deep-agents-code.md): Run interactive and headless Deep Agents Code tasks inside a NemoClaw sandbox. - [Recover and Rebuild Sandboxes](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/operate-sandboxes/recover-and-rebuild-sandboxes.md): Recover a stopped agent runtime or rebuild a sandbox while preserving supported state. - [Update Sandboxes](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/operate-sandboxes/update-sandboxes.md): Update the NemoClaw host CLI and reconcile existing sandboxes with the maintained release. - [Uninstall NemoClaw](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/operate-sandboxes/uninstall-nemoclaw.md): Remove NemoClaw while choosing whether to preserve host-side backups, registry state, OpenShell, and models. - [About Managed MCP Servers](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/mcp-servers/about-managed-mcp-servers.md): Understand how NemoClaw connects sandboxed agents to authenticated Streamable HTTP MCP servers. - [Add an MCP Server](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/mcp-servers/add-an-mcp-server.md): Register an authenticated public or trusted private Streamable HTTP MCP server with a NemoClaw sandbox. - [Manage MCP Servers](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/mcp-servers/manage-mcp-servers.md): Inspect advertised tools and DNS pins, change denied tools, probe, rotate, restart, remove, rebuild, and destroy NemoClaw-managed MCP servers. - [Understand Sandbox State](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/state-and-backups/understand-sandbox-state.md): What agent workspace, sandbox state, and read-only host mounts are, where they live, and how they persist. - [Persist Sandbox State](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/state-and-backups/create-and-restore-snapshots.md): Understand OpenShell native home persistence and NemoClaw's whole-workspace replacement handoff. - [Transfer State Manually](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/manage-sandboxes/state-and-backups/transfer-state-manually.md): Download and upload selected sandbox state files when you need an independent or targeted transfer. - [Approve or Deny Agent Network Requests](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/network-policy/approve-network-requests.md): Review and approve blocked agent network requests in the TUI. - [Customize the Sandbox Network Policy](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/network-policy/customize-network-policy.md): Choose the supported workflow for changing sandbox network access. - [Change the Baseline Network Policy](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/network-policy/configure-policies/change-baseline-network-policy.md): Edit the policy that NemoClaw applies when it creates a sandbox. - [Apply Policy Presets](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/network-policy/configure-policies/apply-policy-presets.md): Add, reapply, list, or remove policy presets for a running NemoClaw sandbox. - [Create Custom Policy Presets](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/network-policy/configure-policies/create-custom-policy-presets.md): Author and apply a scoped network policy preset for an endpoint that NemoClaw does not include. - [Replace the Live Network Policy](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/network-policy/configure-policies/replace-live-network-policy.md): Export, edit, validate, and replace the complete live policy for a NemoClaw sandbox. - [Deploy NemoClaw to a Headless Server](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/deployment/deploy-to-headless-server.md): Install, verify, operate, and recover NemoClaw on a remote Linux server through SSH. - [Register an External Component During Onboarding](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/deployment/register-external-component.md): Register one external host component for validated OpenShell interception and bounded activation during fresh Linux onboarding. - [Declare the OpenShell Gateway Lifecycle Authority](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/deployment/gateway-lifecycle-authority.md): Declare whether NemoClaw or an external platform supervisor owns the OpenShell gateway lifecycle, so one component controls each gateway port. - [Understand Deep Agents Trace Export](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/monitoring/understand-deepagents-trace-export.md): Review the data, redaction limits, transport, and trust boundaries for Deep Agents trace export. - [Set Up Deep Agents Trace Export](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/monitoring/set-up-deepagents-trace-export.md): Enable Deep Agents trace export and configure a host OpenTelemetry collector. - [Verify Deep Agents Trace Export](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/monitoring/verify-deepagents-trace-export.md): Verify Deep Agents trace delivery and diagnose each export hop. - [Manage Deep Agents Trace Export](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/monitoring/manage-deepagents-trace-export.md): Stop, disable, reconfigure, or remove Deep Agents trace export and its host collector. - [NemoClaw Security Posture and Control Trade-Offs](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/security/best-practices.md): Compare NemoClaw security layers and posture profiles, then find the detailed control guidance for each trust boundary. - [Understand Filesystem Controls](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/security/security-controls/filesystem-controls.md): Review NemoClaw filesystem defaults, writable paths, and Landlock enforcement. - [Understand Process Controls](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/security/security-controls/process-controls.md): Review NemoClaw process capabilities, resource limits, runtime identity, and image hardening. - [Understand Gateway and Secret Controls](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/security/security-controls/gateway-authentication-controls.md): Review runtime-specific gateway access and the shared secret-redaction and memory-scanning controls. - [Configure Corporate Certificate Authority Trust](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/security/configure-corporate-ca-trust.md): Import a bounded corporate proxy CA chain into NemoClaw sandbox runtime trust and supported build-time flows. - [Credential Storage](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/security/credential-storage.md): Learn where NemoClaw stores credentials, what protections it applies, and how to inspect or rotate stored secrets. - [Trusted Computing Base for Lifecycle Control](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/security/trusted-computing-base.md): Defines the trusted computing base, privilege boundaries, and review invariants for NemoClaw lifecycle operations. - [Platform Support and Launch Claims](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/platform-support.md): Canonical matrix of supported platforms, inference providers, agents, integrations, and deployment paths for NemoClaw, with explicit status for each row. - [NemoClaw Enterprise Readiness and Admin Capability Guidance](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/enterprise-readiness.md): A reviewed reference for what NemoClaw supports today, what requires manual or admin handling, what is platform-owned, and what is roadmap-only. - [Architecture Details](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/architecture.md): Learn how NemoClaw combines a host CLI, sandbox integration layer, and versioned blueprint to run supported agent runtimes in controlled OpenShell sandboxes. - [Extension Taxonomy and SDK Readiness](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/extension-taxonomy-sdk-readiness.md): Defines NemoClaw extension terminology, stability levels, security boundaries, and measurable readiness gates for any future public SDK. - [NemoDeepAgents CLI Commands Reference](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/commands.md): Full CLI reference for standalone NemoDeepAgents commands and Deep Agents-specific in-sandbox commands. - [Choose Between NemoClaw and OpenShell CLIs](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/cli-selection-guide.md): Choose between the NemoClaw CLI and the OpenShell CLI for common sandbox operations. - [Host Files and State](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/host-files-and-state.md): Reference for NemoClaw host-side state and selected local inference cache paths. - [Network Policies](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/network-policies.md): Baseline network policy, filesystem rules, and operator approval flow. - [Troubleshoot MCP Servers](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/troubleshoot-mcp-servers.md): Diagnose managed MCP transport failures, timeout evidence, credential resolution, DNS pin drift, incomplete transactions, capability gaps, policy drift, and lifecycle-lock failures. - [System Readiness](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/system-readiness.md): Check host and gateway readiness before NemoClaw onboarding without changing system state. - [Troubleshooting](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/reference/troubleshooting.md): Diagnose and resolve common NemoClaw installation, onboarding, and runtime issues. - [Use NemoClaw Docs with Your Coding Agents](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/resources/agent-skills.md): NemoClaw exposes an MCP docs server and Markdown documentation that AI coding agents can use when they help you install, configure, or operate a sandbox. - [NemoClaw Community Solutions](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/resources/community-contributions.md): Browse examples or choose where to contribute a solution. - [Engineer Documentation for AI Agents](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/resources/engineer-agentic-documentation.md): A practical architecture and operating model for making technical documentation reliable for both people and AI agents. - [License](https://docs.nvidia.com/nemoclaw/latest/user-guide/deepagents/resources/license.md): Apache 2.0 license for the NemoClaw project.