GSC-MCP-Server
Google Search Console MCP Server — Connect Google Search Console to Claude, Cursor, and other MCP clients.
接入你的智能体
從發布者安裝。不是 PlugMCP 端點。
遠端 URL 還是本機指令? 兩種安裝方式的差異
# Install globally
npm install -g @appsyogi/gsc-mcp-server
# Set up OAuth credentials
gsc-mcp init
# Verify setup
gsc-mcp doctor
# Start the server (for MCP clients)
gsc-mcp run
You need to create your own OAuth credentials in Google Cloud Console:
Note: While your app's publishing status is "Testing", only test users can authenticate. You can add up to 100 test users.
Run the init command and enter your credentials:
gsc-mcp init
This will:
# Initialize with OAuth (interactive)
gsc-mcp init
# Initialize with service account
gsc-mcp init --service-account /path/to/key.json
# Initialize with full scope (for sitemap submission, URL inspection)
gsc-mcp init --scope full
# Check configuration and connectivity
gsc-mcp doctor
gsc-mcp doctor --verbose
# Start MCP server (stdio mode)
gsc-mcp run
# Start in HTTP mode (for debugging)
gsc-mcp run --http 3333
# View current configuration
gsc-mcp config
# Clear credentials
gsc-mcp logout
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
{
"mcpServers": {
"gsc": {
"command": "gsc-mcp",
"args": ["run"]
}
}
}
Add to your VS Code MCP settings (~/.vscode/mcp.json or workspace settings):
{
"servers": {
"gsc": {
"command": "gsc-mcp",
"args": ["run"],
"type": "stdio"
}
}
}
Add to your Cursor MCP config:
{
"mcpServers": {
"gsc": {
"command": "npx",
"args": ["-y", "@appsyogi/gsc-mcp-server", "run"]
}
}
}
All tools support parameters to reduce token consumption for LLM use cases:
Add format: "compact" to any tool for LLM-optimized output:
{
"siteUrl": "sc-domain:example.com",
"startDate": "2026-01-01",
"endDate": "2026-01-28",
"dimensions": ["query"],
"format": "compact"
}
Compact format features:
imp instead of impressions, pos instead of position)"5.41%" instead of 0.054123...)Example compact response:
{
"summary": "Top query 'example keyword' got 150 clicks at position 3.2",
"rows": [
{"key": "example keyword", "clicks": 150, "imp": 2800, "ctr": "5.36%", "pos": 3.2}
]
}
For date dimension queries, use granularity to reduce row count:
{
"siteUrl": "sc-domain:example.com",
"startDate": "2025-07-01",
"endDate": "2026-01-28",
"dimensions": ["date"],
"granularity": "weekly"
}
Options:
daily — Default, returns daily dataweekly — Aggregates by week (Monday-based)monthly — Aggregates by monthauto — Picks based on date range (>90 days = monthly, >21 days = weekly)Default rowLimit is 25 (optimized for LLM context windows). Use higher values when needed:
{
"siteUrl": "sc-domain:example.com",
"rowLimit": 100
}
Maximum: 25,000 rows.
The server exposes browsable resources:
gsc://sites — List all propertiesgsc://sites/{siteUrl}/sitemaps — List sitemaps for a propertyGSC-MCP supports two permission levels:
gsc-mcp init # Uses readonly by default
gsc-mcp init --scope full
GSC-MCP stores configuration in platform-specific locations:
Files:
config.json — OAuth client ID/secret, scope settingscache.sqlite — Query cache and saved presetsTokens are stored securely in the OS keychain when available.
For automated/server use, you can use a service account instead of OAuth:
gsc-mcp init --service-account /path/to/key.json
Google Search Console API has a default quota of 1,200 queries per day. GSC-MCP includes:
Use gsc-mcp to find quick-win opportunities for my site https://example.com
Generate a weekly performance summary for https://example.com
Compare search performance for https://example.com between last week and the week before
Export the top 1000 queries for https://example.com in the last 28 days as CSV
# Clone the repo
git clone https://github.com/AppsYogi-com/gsc-mcp-server.git
cd gsc-mcp-server
# Install dependencies
npm install
# Build
npm run build
# Run in dev mode
npm run dev
# Test locally
node dist/cli/index.js doctor
MIT
Contributions are welcome! Please open an issue or PR.
Built with:
Google Search Console MCP Server - Connect Google Search Console to Claude, Cursor, and other MCP clients.
工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式,而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
Gsc 透過 stdio 在本機執行。
上次檢查時間:2026-11-12。 Google Search Console MCP Server - Connect Google Search Console to Claude, Cursor, and other MCP clients.