Scrapling Fetch MCP

from mcp.types import Tool from pydantic import BaseModel, Field class PageFetchRequest(BaseModel): url: str = Field(..., description="URL to fetch") mode: str = Field( "basic", description="Fetching mode (basic, stealth, or max-stealth)" ) format: str = Field("markdown", description="Output format (html or markdown)") max_length: int = Field( 5000, description="Maximum number of characters to return.", gt=0, lt=1000000, title="Max Length", ) start_index: int = Field( 0, description="On return output starting at this character index, useful if a previous fetch was truncated and more content is required.", ge=0, title="Start Index", ) class PatternFetchRequest(BaseModel): url: str = Field(..., description="URL to fetch") mode: str = Field( "basic", description="Fetching mode (basic, stealth, or max-stealth)" ) format: str = Field("markdown", description="Output format (html or markdown)") max_length: int = Field( 5000, description="Maximum number of characters to return.", gt=0, lt=1000000, title="Max Length", ) search_pattern: str = Field( ..., description="Regular expression pattern to search for in the content", ) context_chars: int = Field( 200, description="Number of characters to include before and after each match", ge=0, ) s_fetch_page_tool = Tool( name="s-fetch-page", description="Fetches a complete web page with pagination support. " "Retrieves content from websites with bot-detection avoidance. " "For best performance, start with 'basic' mode (fastest), then only escalate to " "'stealth' or 'max-stealth' modes if basic mode fails. " "Content is returned as 'METADATA: {json}\\n\\n[content]' where metadata includes " "length information and truncation status.", inputSchema=PageFetchRequest.model_json_schema(), ) s_fetch_pattern_tool = Tool( name="s-fetch-pattern", description="Extracts content matching regex patterns from web pages. " "Retrieves specific content from websites with bot-detection avoidance. " "For best performance, start with 'basic' mode (fastest), then only escalate to " "'stealth' or 'max-stealth' modes if basic mode fails. " "Returns matched content as 'METADATA: {json}\\n\\n[content]' where metadata includes " "match statistics and truncation information. Each matched content chunk is " "delimited with '॥๛॥' and prefixed with '[Position: start-end]' indicating its byte position " "in the original document, allowing targeted follow-up requests with s-fetch-page using " "specific start_index values.", inputSchema=PatternFetchRequest.model_json_schema(), )