LINKTH CRAWL MCP Server
linkth-crawl-mcp lets MCP clients call the hosted LINKTH CRAWL SEO crawler API over stdio.
The server is a thin client. It does not bundle secrets, does not connect to your database, and only calls the public hosted API with your per-user uck_live_ API key.
Requirements
- Node.js 18 or newer
- A LINKTH CRAWL API key from the app dashboard
- Hosted API base URL:
https://crawler.link-th.co.jp
Usage
LINKTH_API_KEY=uck_live_xxx npx -y linkth-crawl-mcp
Optional:
LINKTH_API_BASE=https://crawler.link-th.co.jp
Backward-compatible environment variables are still accepted:
LINKSCRAWL_API_KEYLINKSCRAWL_API_BASE
New integrations should use LINKTH_API_KEY and LINKTH_API_BASE.
MCP Client Config
.mcp.json example:
{
"mcpServers": {
"linkth-crawl": {
"command": "npx",
"args": ["-y", "linkth-crawl-mcp"],
"env": {
"LINKTH_API_KEY": "uck_live_xxx",
"LINKTH_API_BASE": "https://crawler.link-th.co.jp"
}
}
}
}
Supported Tools
These tools are exposed by the public MCP server:
| Tool | Purpose |
|---|---|
list_jobs |
List your crawl jobs. |
crawl_site |
Start a crawl. |
crawl_status |
Check crawl progress. |
list_pages |
List crawled pages. |
get_links |
Inspect internal link edges. |
orphans |
Find orphan pages. |
structure_summary |
Summarize site structure. |
page_detail |
Inspect one crawled page. |
find_similar |
Find similar pages. |
pagerank |
Return PageRank scores. |
issues |
Return detected SEO issues. |
extract is a paid/admin-only hosted API feature and is intentionally excluded from the public beta MCP tool list.
generate_proposals is intentionally not exposed in this package because the hosted endpoint currently returns 501 until behavior-data tenant mapping is wired.
Development
npm install
npm run build
npm test
npm run typecheck
Run locally:
LINKTH_API_KEY=uck_live_xxx LINKTH_API_BASE=https://crawler.link-th.co.jp node dist/index.js
Publish
The package is prepared for public npm publishing, but publishing must be done by the package owner:
npm login
npm publish --access public