What is pure.md?

pure.md provides a robust REST API infrastructure that enables AI agents and developers to interact seamlessly with web content. It acts as a global cache, facilitating reliable access to websites while minimizing the risk of being blocked. The service is engineered to mimic real user behavior, utilizing rotating IP addresses and sophisticated browser fingerprint emulation to circumvent bot detection systems. If direct access fails, it automatically attempts retrieval from Common Crawl and Internet Archive datasets as fallbacks.

The platform excels at rendering dynamic web content, including JavaScript-heavy single-page applications (SPAs), ensuring complete page rendering where simple fetching methods would fail. It also processes various file types, converting PDFs, images (with AI-driven object detection and summarization), and spreadsheet documents into markdown format. This markdown output is specifically optimized for Large Language Models (LLMs), stripping superfluous elements and adding relevant metadata to provide maximum context with minimal token usage, thereby reducing inference costs and improving workflow speed for AI applications.

Features

Bot Detection Avoidance: Mimics real browser fingerprints and rotates IPs to prevent being flagged.
Fallback Data Sources: Seamlessly uses Common Crawl and Internet Archive if direct access fails.
Headless Content Rendering: Renders JavaScript-heavy SPAs, PDFs, images, and spreadsheets.
LLM-Optimized Markdown: Converts web pages and files into low-token, context-rich markdown.
SERP Crawling: Fetches and concatenates search engine results for real-time knowledge.
Natural Language Data Extraction: Extracts structured (JSON) or unstructured data using AI models and prompts.
AI Model Selection: Offers various generative AI models (e.g., Llama, Mistral) for extraction tasks.
MCP Server Support: Compatible with Model Context Protocol for integration with tools like Cursor.

Use Cases

Enhancing AI agents with reliable web access.
Scraping web data for training or feeding LLMs.
Bypassing bot detection measures for web crawling.
Extracting structured data from websites using natural language queries.
Providing AI applications with real-time information from search engines.
Converting diverse web content (HTML, PDF, images) into a unified markdown format.

FAQs

Do I need a credit card to sign up for pure.md?

No, you can sign up without a credit card, but you will have a strict rate limit until you activate a subscription.
How do the free credits work on pure.md plans?

You pay a flat fee upfront (except on Starter). Your usage deducts from these credits. Once credits are used, you're billed based on usage. Unused credits don't roll over.
How much does data extraction cost with pure.md?

Data extraction pricing varies by the generative AI model used and is based on input and output tokens. Costs apply only to POST requests, not GET requests.
Can pure.md access content behind a login?

Yes, include your authorization cookies in the request headers, and pure.md will pass them along to the target URL.
What file types does pure.md support for markdown conversion?

HTML (.htm, .html, .xml), PDF (.pdf), images (.jpeg, .jpg, .png, .svg, .webp), and spreadsheet files (.csv, .et, .numbers, .xls, .xlsb, .xlsm, .xlsx) are supported.