Mcp Server
Let Claude or Cursor fetch any public URL and return structured JSON through the Haunt API instead of brittle manual copy-paste.
Overview
Haunt MCP is a MCP server for the Build phase that converts any URL into structured JSON for Claude and Cursor via the Haunt API.
What is this MCP server?
- npm @hauntapi/mcp-server version 1.0.3 with stdio transport
- Turns arbitrary URLs into structured JSON for agent pipelines
- Requires HAUNT_API_KEY secret from hauntapi.com per server manifest
- Positioned for Claude and Cursor workflows in the registry description
- GitHub source at Darko893/mcp-server for install and issue tracking
- Package version 1.0.3
- One required secret environment variable: HAUNT_API_KEY
- stdio npm identifier @hauntapi/mcp-server
What problem does it solve?
Agents stall on messy HTML and ad-hoc scraping when you need clean JSON from live web pages.
Who is it for?
Indie builders automating research, monitoring, or content ingestion inside Claude Code or Cursor.
Skip if: Workflows that need logged-in sessions, heavy JavaScript SPAs without API support, or zero third-party API budget.
What do I get? / Deliverables
After you set HAUNT_API_KEY and register the stdio server, your agent extracts structured page data in one tool call.
- Structured JSON payloads from submitted URLs
- Agent-callable extraction tools in the Haunt MCP surface
- Repeatable web ingest step for research and build pipelines
Recommended MCP Servers
Journey fit
Web extraction MCP is most often wired while building automations, datasets, or agent tools that consume live pages. integrations is correct because the Haunt MCP wraps a third-party extraction API behind MCP tools with a required API key.
How it compares
Haunt API web-extraction MCP, not local repo memory or nutrition data services.
Common Questions / FAQ
Who is Haunt MCP for?
Solo developers and agent users who need dependable URL-to-JSON extraction inside MCP-aware editors without maintaining custom scrapers.
When should I use Haunt MCP?
Use it during build or research when your agent must read public web pages as structured data for analysis, enrichment, or prototyping.
How do I add Haunt MCP to my agent?
Install @hauntapi/mcp-server, add HAUNT_API_KEY to the MCP server environment in Claude Code or Cursor, and register the package as a stdio server.