agentskills.codes

--- name: mcporter description: Call MCP servers via MCPorter - browser automation, vision, and other MCP tools homepage: https://github.com/steipete/mcporter metadata: { "openclaw": { "emoji": "🧳", "requires": { "bins": ["npx"] } } }

Install

mkdir -p .claude/skills/mcporter-consuelohq && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/16048" && unzip -o skill.zip -d .claude/skills/mcporter-consuelohq && rm skill.zip

Installs to .claude/skills/mcporter-consuelohq

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

--- name: mcporter description: Call MCP servers via MCPorter - browser automation, vision, and other MCP tools homepage: https://github.com/steipete/mcporter metadata: { "openclaw": { "emoji": "🧳", "requires": { "bins": ["npx"] } } }
235 chars · catalog descriptionno explicit “when” trigger

About this skill


name: mcporter description: Call MCP servers via MCPorter - browser automation, vision, and other MCP tools homepage: https://github.com/steipete/mcporter metadata: { "openclaw": { "emoji": "🧳", "requires": { "bins": ["npx"] } } }

MCPorter

MCPorter is a TypeScript runtime and CLI that bridges OpenClaw to Model Context Protocol (MCP) servers. Use this skill to:

  • Browser automation via chrome-devtools MCP (snapshots, clicks, forms, navigation, screenshots)
  • Vision analysis via Z.AI MCP server (images, videos)
  • Code automation via Playwright MCP
  • Any other MCP server configured in MCPorter

Quick Start

List available MCP servers and tools

npx mcporter list

Call a specific MCP tool

# Chrome DevTools examples (✅ working - tested with GitHub.com)
npx mcporter call chrome-devtools.take_snapshot
npx mcporter call chrome-devtools.navigate_page type:url url:"https://example.com"
npx mcporter call chrome-devtools.take_screenshot format:png
npx mcporter call chrome-devtools.click uid:"element-id"
npx mcporter call chrome-devtools.fill uid:"input-id" value:"hello world"
npx mcporter call chrome-devtools.evaluate_script function:"() => document.title"
npx mcporter call chrome-devtools.list_pages
npx mcporter call chrome-devtools.list_console_messages
npx mcporter call chrome-devtools.list_network_requests

# Z.AI Vision examples (⚠️ requires Z_AI_API_KEY)
npx mcporter call zai-vision.image_analysis filePath:"/path/to/image.png"
npx mcporter call zai-vision.video_analysis filePath:"/path/to/video.mp4"

Key MCP Servers

chrome-devtools (26 tools)

Browser automation and inspection:

  • take_snapshot — Page snapshot (a11y tree with element UIDs)
  • click — Click element by UID
  • hover — Hover over element
  • drag — Drag element to drop target
  • fill — Fill form input
  • fill_form — Fill multiple form elements
  • navigate_page — Navigate (url/back/forward/reload)
  • new_page — Open new page
  • select_page — Select page context
  • close_page — Close page
  • take_screenshot — Screenshot (PNG/JPEG/WebP)
  • evaluate_script — Run JavaScript
  • press_key — Keyboard input
  • resize_page — Resize viewport
  • emulate — Network/CPU/geolocation/user agent emulation
  • list_console_messages — Console logs
  • list_network_requests — Network logs
  • get_console_message — Get specific console message
  • get_network_request — Get specific network request
  • upload_file — Upload via file input
  • handle_dialog — Handle browser dialogs
  • wait_for — Wait for text/element
  • performance_start_trace / performance_stop_trace — Performance tracing
  • performance_analyze_insight — Analyze performance insights

microsoft/playwright-mcp (22 tools)

Browser automation via Playwright (currently has connection issues).

zai-vision (requires Z_AI_API_KEY)

Image and video analysis:

  • ui_to_artifact — UI screenshots to code/specs/prompts
  • extract_text_from_screenshot — OCR and text extraction from screenshots
  • diagnose_error_screenshot — Analyze error screenshots and provide solutions
  • understand_technical_diagram — Diagram interpretation (architecture, flow, UML, ER)
  • analyze_data_visualization — Charts and dashboard analysis
  • ui_diff_check — Compare two UIs to identify differences
  • image_analysis — General-purpose image understanding (fallback)
  • video_analysis — Video understanding (≤8MB, MP4/MOV/M4V)

upstash/context7 (2 tools)

Library documentation:

  • resolve_library_id — Get library ID
  • get_library_docs — Fetch library docs

Configuration

View current MCP servers

npx mcporter list

Add a new MCP server

npx mcporter config add <name> <target>

# Examples:
npx mcporter config add zai-vision "npx -y @z_ai/mcp-server"
npx mcporter config add chrome-devtools "npx -y chrome-devtools-mcp@latest"

Set API keys for MCP servers that need them

export Z_AI_API_KEY=your-key-here

# Or add to shell profile (~/.zshrc, ~/.bashrc, ~/.config/fish/config.fish)
echo 'export Z_AI_API_KEY=your-key-here' >> ~/.zshrc

Notes

MCPorter auto-discovers MCP servers from:

  • Cursor (~/.cursor/mcp.json)
  • Claude Desktop (~/.claude.json)
  • Claude Code (~/.claude-desktop/mcp.json)
  • Codex (~/Library/Application Support/Code/User/mcp.json)
  • VS Code
  • Windsurf
  • OpenCode

Workspace config

OpenClaw's MCPorter config lives at: /Users/kokayi/.openclaw/workspace/config/mcporter.json

Current configured servers:

  • chrome-devtools (26 tools) — ✅ Working (via stdio)
  • zai-vision (8 tools) — ⚠️ Needs Z_AI_API_KEY

Auto-discovered servers

The following MCP servers are auto-discovered from ~/.claude.json:

  • chrome-devtools (from Claude Desktop) — 26 tools, browser automation
  • upstash/context7 (from Code) — 2 tools, library docs

Note: Auto-discovered servers are available via their source tools (e.g., Claude Desktop, VS Code) but may not be directly accessible through MCPorter CLI without importing them to the workspace config.

When to use MCP tools

  • Browser automation — Use chrome-devtools MCP tools
  • Vision/OCR — Use zai-vision MCP tools (requires API key)
  • Complex workflows — Combine multiple MCP calls

Syntax variations (all work)

# Function-call style (recommended)
npx mcporter call 'chrome-devtools.click(uid: "element-id")'

# Flag style
npx mcporter call chrome-devtools.click uid:"element-id"

# Colon-delimited
npx mcporter call chrome-devtools.click uid:"element-id"

Troubleshooting

"tools unavailable" error

If you see "tools unavailable" for a server, it means:

  • MCP server is not running (for stdio)
  • Connection failed (for HTTP)
  • Missing API keys or auth

Fix for microsoft/playwright-mcp

Currently shows SSE/connection errors. This is a known issue with the MCP server.

Z.AI MCP requires API key

Set environment variable:

export Z_AI_API_KEY=your-zai-api-key

Or add to MCPorter config (already configured ✅):

{
  "mcpServers": {
    "zai-vision": {
      "command": "npx -y @z_ai/mcp-server",
      "env": {
        "Z_AI_API_KEY": "your-zai-api-key"
      }
    }
  }
}

✅ Configured: Z.AI API key is set in MCPorter config. Vision tools are ready to use!

Resources

Search skills

Search the agent skills registry