windows-app-mcp

Inspect and invoke the OpenClaw Windows app through its local MCP server and winnode CLI.

2.1k|275|Updated Jan 29, 2026
One-click install
npx skills add https://github.com/openclaw/openclaw-windows-node --skill windows-app-mcp
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: windows-app-mcp
Source: https://github.com/openclaw/openclaw-windows-node/tree/main/.agents/skills/windows-app-mcp
Command: npx skills add https://github.com/openclaw/openclaw-windows-node --skill windows-app-mcp

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Developers working on the OpenClaw Windows app need a deterministic way to inspect, invoke, and validate the running app's capabilities without going through the full gateway pairing flow, and without confusing local MCP proof with gateway-mediated behavior.

Core Features & Use Cases

  • Local MCP discovery and invocation: Run winnode --list-tools to discover the live tool surface and invoke commands like app.status, system.which, or screen.* with JSON parameters against the authenticated loopback MCP server.
  • Isolated development environment: Launch the app with run-app-local.ps1 -Isolated -AllowNonMain so tests never touch the real %APPDATA%\OpenClawTray data, and run MCP-only mode without gateway credentials.
  • Raw protocol proof: Send raw JSON-RPC initialize, tools/list, and tools/call messages to validate server or protocol changes, including cancellation via notifications/cancelled.
  • Use Case: After adding a new Windows node command, register it through the shared INodeCapability path, update McpToolBridge.CommandDescriptions and the CLI reference, then prove it end-to-end with winnode --list-tools and winnode --command before validating the gateway path separately.

Quick Start

Ask the agent to start an isolated OpenClaw Windows app instance, list the available local MCP tools with winnode, and invoke the app.status command to verify the connection.

Frequently Asked Questions about windows-app-mcp

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I list and invoke local MCP tools in the OpenClaw Windows app?

Run `winnode --list-tools` to discover the live tool surface, then invoke commands with `winnode --command app.status --params '{}'`. The live tools/list response is authoritative because settings determine which capabilities are exposed.

How do I run the OpenClaw Windows app in an isolated test environment?

Launch it with `run-app-local.ps1 -Isolated -AllowNonMain`, then set OPENCLAW_TRAY_DATA_DIR to the printed isolated data directory. This keeps tests from modifying the real %APPDATA%\OpenClawTray data.

Can I use the local MCP server without a gateway connection?

Yes, MCP-only mode is supported by setting EnableMcpServer=true with EnableNodeMode=false, which starts local capabilities without requiring gateway credentials. Enable the Local MCP Server option in Permissions first.

What is the difference between local MCP and OpenClaw node mode?

Node mode is the product path where agents reach a paired Windows machine through the gateway WebSocket and node.invoke. Local MCP is an authenticated loopback HTTP surface for same-machine developers, and local MCP results do not prove gateway behavior.

Why should I not put the MCP token on the winnode command line?

winnode loads the MCP token automatically from the selected tray profile, so passing it manually is unnecessary and risks exposing the bearer token. Never publish the token, settings, gateway records, or secret-bearing responses.

What steps are required when adding a new Windows node command?

Register it through the shared INodeCapability path, update McpToolBridge.CommandDescriptions and the CLI skill reference, add capability and winnode tests, then prove it with winnode --list-tools and a command invocation. Validate the gateway path separately if gateway behavior changed.