Skip to content

Add TheCrawler as example MCP tool — web extraction with structured errors #692

Description

@Rashadamom

Tool Example Request

TheCrawler would be a useful example MCP tool for mcp-agent workflows that need web data but also need programmatic failure handling.

What it does:

  • URL → markdown via crawl_markdown
  • Full crawl output via crawl with metadata, links, images, JSON-LD, microdata, commerce/forms/analytics signals, hreflang, pagination, and redirect chain
  • PDF/DOCX parsing
  • Structured errors (blocked-bot, timeout, http-4xx, js-required, etc.) so agents can branch instead of trusting empty output
  • Optional schema-backed structured extraction through OpenAI-compatible LLM endpoints
  • Current GitHub source includes validated extraction contract diagnostics for checking whether a URL set is extract-ready

Quick start from GitHub source:

git clone https://github.com/manchittlab/TheCrawler.git
cd TheCrawler/engine
npm install
npm run build
node dist/mcp.js

MCP engine path: https://github.com/manchittlab/TheCrawler/tree/main/engine
License: AGPL-3.0-or-later

Note: npm currently lags the GitHub source, so the GitHub-source path is the current install path for validated-contract diagnostics.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions