AntonG87/codearia-sieve

Overview

Parses web pages into the exact format a decision model needs, converting dates into date objects, numbers into values with units, and splitting content into chunks that fit the model's context window. Available as an MCP server and a TypeScript library. No model or API key is required.

Best for

Best for
Developers preparing web content for decision models without relying on external AI services.

Use cases

• Extract dates and unit-bearing numbers from web pages for model input
• Chunk long web content to fit a decision model's context window
• Normalize web page data into structured fields before analysis

How to use

Install

npx codearia-sieve          # MCP server for Claude Code, Cursor and any agent

Tools exposed

• robots-disallowed
• http-error
• empty-without-js
• thin-content
• block-split
• facts-capped

Tested with

Claude Code, Cursor

Notes

Parses web pages into the exact format a decision model needs, converting dates into date objects, numbers into values with units, and splitting content into chunks that fit the model’s context window. Available as an MCP server and a TypeScript library. No model or API key is required.

5 stars on GitHub. Last updated 2026-09-22. Licensed MIT.

Use cases

• Extract dates and unit-bearing numbers from web pages for model input
• Chunk long web content to fit a decision model’s context window
• Normalize web page data into structured fields before analysis

Pros

• No API key or model dependency required
• Dual distribution as MCP server and TypeScript library
• Handles date and unit normalization in one step

Cons

• Only 5 GitHub stars, indicating limited adoption
• Requires self-hosting the MCP server or integrating the TypeScript library
• Focused only on parsing, not on model inference or downstream processing

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.