# Jina Reader

> Converts any URL or web search query into clean, LLM-friendly Markdown by prepending r.jina.ai or s.jina.ai.

Reader is a Jina AI API, with an open-source branch of its codebase on GitHub, that turns web pages into LLM-friendly input for agent and RAG systems. Prepending `r.jina.ai` to a URL returns cleaned content, and `s.jina.ai` searches the web and returns the content of the top results. The repository is licensed under Apache-2.0 and can be self-hosted with Docker.

## What It Is

Reader is developer infrastructure for grounding LLMs with web content. It fetches a URL server-side, strips boilerplate such as navigation and ads, and converts the main content to Markdown. Jina's documentation states it is not a consumer search engine and does not index or rank the web. Users control which URLs are processed.

## Read and Search Workflows

The read endpoint handles web pages, PDFs, MS Office documents and images. Pages are rendered with headless Chrome or fetched through a lightweight curl path. Images can be captioned by a vision-language model so text-only LLMs get hints about visuals. The search endpoint fetches the top five results and runs each through the same pipeline, with optional site restriction. Documents such as PDFs and Office files can be POSTed directly.

## Controlling the Output

Request headers control the behavior of the API:
- Output format: markdown, html, text, screenshot, pageshot, or frontmatter.
- Fetching engine selection.
- CSS target, wait and remove selectors.
- Token limits and budgets.
- Proxy and cookie forwarding.
- Caching controls.
- Link and image handling.
- Semantic chunking.
- Presets for reader, index, research, agent and spider use.

JSON responses and a streaming mode are also supported. Structured extraction with a JSON schema or natural-language instruction uses the ReaderLM-v2 model.

## Deployment Options

The open-source branch runs statelessly out of the box, with optional S3-compatible bucket caching, and a prebuilt Docker image is published to GitHub Container Registry. Jina says the hosted service stays available through the API, and an MCP server is offered at mcp.jina.ai. The models used by Reader, such as ReaderLM-v2 and jina-vlm, are CC-BY-NC 4.0 licensed, with commercial licensing sold through Elastic.

## Features
- Convert any URL to LLM-friendly Markdown via r.jina.ai prefix
- Web search returning top 5 results with full content via s.jina.ai
- PDF and MS Office document parsing
- Image captioning with a vision-language model
- Selectable fetching engines (browser, curl, auto)
- CSS target, wait and exclude selectors
- JSON and streaming response modes
- Structured extraction with JSON schema or instructions via ReaderLM-v2
- Proxy, cookie forwarding and caching controls
- Semantic markdown chunking and presets
- Self-hosting with Docker
- MCP server access

## Integrations
Model Context Protocol, Puppeteer, Docker, MinIO, Google Colab, AWS SageMaker, Google Cloud Marketplace, Microsoft Azure Marketplace

## Platforms
MACOS, WEB, API, VSC_EXTENSION, CLI

## Pricing
Open Source

## Links
- Website: https://jina.ai/reader
- Documentation: https://r.jina.ai/docs
- Repository: https://github.com/jina-ai/reader
- EveryDev.ai: https://www.everydev.ai/tools/jina-reader