Node Command

Operate your Windows, macOS, and Linux machines from any AI assistant — through one Model Context Protocol (MCP) endpoint.

Give ChatGPT, Claude, Grok, Muse, and Gemini direct, secure access to your computers. Read files, inspect services, and run commands over outbound-only connections with strict policy ceilings and a cryptographic audit trail.

• One connector URL — https://mcp.nodecommand.app/mcp. Clients that support MCP OAuth discovery sign in automatically; nothing to copy or paste.
• Single-command install — npx node-command pairs the machine with an eight-character code, installs it as a service, and verifies it online.
• Multi-host fleets — talk to a cloud VPS, a home lab server, and a laptop in the same conversation. No inbound ports, no firewall changes.
• Team access — share machines as Viewer, Operator, or Admin. Never share SSH keys or credentials again.
• Cryptographic proof — every action is signed and chained in an immutable SHA-256 audit trail. Know who ran what, when, on which machine, and with what result.
• Hard safety limits — per-machine policy ceilings restrict file access, block dangerous commands, and enforce read-only execution modes.

Quick start

1. Install the agent on the machine you want to control:

   npx node-command

   Approve the pairing in your browser with the eight-character code shown in your terminal. The agent connects outbound-only and runs as a system or user service. For hands-off enrollment, redeem the dashboard's one-time enrollment token directly with npx node-command <host_id> <token>. The previous npx nc-agent command remains supported as an alias.

2. Connect your AI assistant. Add the MCP endpoint above as a custom connector or MCP server. Ready-made configurations live in examples/ for ChatGPT, Claude, Cursor, VS Code, and OpenCode. Sign-in is automatic for clients that support MCP OAuth; header-based clients can use a personal access token.

3. Ask. Your assistant can see every enrolled machine in your workspace, and host actions always target an explicit machine — the platform never guesses one.

Full walkthrough: docs/getting-started.md.

Documentation

Guide	What it covers
Getting started	Account, first machine, first connection
Agent installation	Install, verify, update, uninstall
Architecture	Hub, agent, and contract — what is open
Security model	Tokens, device identity, audit chain, policy
Clients (hosts)	Machine IDs, fleet views, explicit targeting
Troubleshooting	Common symptoms and fixes

Related repositories

• nc-agent — the open-source host agent and CLI installer that runs on your machines.
• nc-protocol — the canonical public wire contracts: agent protocol and MCP tool schemas.

Contributing, security, license

• Contributions: CONTRIBUTING.md — docs, examples, and manifests live here; code lives in the two repositories above.
• Vulnerabilities: SECURITY.md — coordinated disclosure, never a public issue.
• License: Apache-2.0 — LICENSE. Brand usage: TRADEMARKS.md.