pure.md is a REST API that lets AI agents and developers reliably access web content. With pure.md, you can:
Generate a unique API token in your dashboard.
Then include that token in the x-puremd-api-token
request header for all requests.
Rate limits vary by subscription plan. See pricing for details.
Subscription type | Requests per minute |
---|---|
Logged out / anonymous | 6 |
Logged in, no subscription | 10 |
Starter plan | 60 |
Growth plan | 600 |
Business plan | 3000 |
The Model Context Protocol, developed by Anthropic, is an open standard that enables AI systems to seamlessly interact with an ecosystem of tooling. With it, MCP clients like Cursor, Windsurf, and Claude Desktop can learn how to use a variety of APIs and other functionality.
You can instruct your MCP clients to route traffic through pure.md by following the instructions at https://github.com/puremd/puremd-mcp.
All request headers pass through to the target URL, except ones that begin with x-puremd-
.
Original headers from the origin are returned in the response.
Retrieves the content of a given URL in markdown format. Use this endpoint to scrape text content from a web page without getting blocked.
url required | string The URL |
<WebPage url="https://example.com"> title: Example Domain access_date: Wed, 05 Mar 2025 22:27:19 GMT --- # Example Domain This domain is for use in illustrative examples in documents. You may use this domain in literature without prior coordination or asking for permission. More information... </WebPage>
This endpoint is only available on paid plans.
Runs inference on the content of a given URL. Use this endpoint to extract structured JSON from a webpage.
url required | string The URL |
prompt required | string The user message |
model | string Enum: "meta/llama-3.1-8b" … 3 more The generative AI model to use. Smaller models are faster, while larger models are more accurate. Default model: |
schema | object JSON schema of the desired response. Omit this property to get a response in plaintext. |
{- "prompt": "What are the top 5 headlines from today?",
- "model": "meta/llama-3.1-8b",
- "schema": {
- "type": "object",
- "properties": {
- "headlines": {
- "type": "array",
- "items": {
- "type": "string"
}
}
}, - "required": [
- "headlines"
]
}
}
{- "type": "object",
- "properties": {
- "headlines": {
- "type": "array",
- "items": {
- "type": "string"
}
}
}, - "required": [
- "headlines"
]
}
This endpoint is only available on paid plans.
Crawls the top results from a search engine query and concatenates the web content from all pages into markdown. Use this endpoint to gather knowledge of news, current events, or specific topics.
q required | string The URL-encoded search query |
# Title of the Page ## Introduction This is the introduction text from the webpage, purified and optimized for LLM processing. ## Main Content The main content of the page converted to clean markdown format, with unnecessary elements removed. ### Subsection Content organized in logical subsections with proper hierarchy. ## Conclusion The concluding information from the webpage.
This endpoint is only available on paid plans.
Crawls the top results from a search engine query and runs inference on their web content. Use this endpoint to answer questions about news, current events, or general user queries that require searching.
url required | string The URL |
prompt required | string The user message |
model | string Enum: "meta/llama-3.1-8b" … 3 more The generative AI model to use. Smaller models are faster, while larger models are more accurate. Default model: |
schema | object JSON schema of the desired response. Omit this property to get a response in plaintext. |
{- "prompt": "Who won the baseball game last night?",
- "model": "meta/llama-3.1-8b",
- "schema": {
- "type": "object",
- "properties": {
- "headlines": {
- "type": "array",
- "items": {
- "type": "string"
}
}
}, - "required": [
- "headlines"
]
}
}
"string"