OCR for agents: PDFs and images to Markdown or JSON, routed across OCR models. Free models included.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Bearer <your OCRRouter API key> from https://ocrrouter.ai/keys
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
List the OCR models you can route to, with credits per page (0 = free) and whether each is available.
This tool takes no arguments.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add --transport http ocrrouter https://api.ocrrouter.ai/v1/mcp \ --header 'Authorization: Bearer YOUR_TOKEN'
Headers
Authorization (required): Bearer <your OCRRouter API key> from https://ocrrouter.ai/keysList the OCR models you can route to, with credits per page (0 = free) and whether each is available.
No arguments.
{
"type": "object",
"properties": {}
}Read a PDF or image and return its text as Markdown (tables kept as tables) or JSON blocks. Pass file_url for a public link, or file_base64 for file contents. Models jina/jina-ocr and deepseek/deepseek-ocr are free; auto picks the best model for the document.
| Argument | Type | Description |
|---|---|---|
| model | string | Model ID from list_models, or auto (default). |
| pages | string | Page range such as 1-5,8. Default all pages. |
| format | string | Output format. Default markdown. |
| file_url | string | Public http(s) URL of a PDF, PNG, JPG, WebP or GIF. |
| filename | string | File name with extension, recommended with file_base64. |
| file_base64 | string | Base64 file contents, instead of file_url. |
{
"type": "object",
"properties": {
"model": {
"type": "string",
"description": "Model ID from list_models, or auto (default)."
},
"pages": {
"type": "string",
"description": "Page range such as 1-5,8. Default all pages."
},
"format": {
"enum": [
"markdown",
"json",
"text"
],
"type": "string",
"description": "Output format. Default markdown."
},
"file_url": {
"type": "string",
"description": "Public http(s) URL of a PDF, PNG, JPG, WebP or GIF."
},
"filename": {
"type": "string",
"description": "File name with extension, recommended with file_base64."
},
"file_base64": {
"type": "string",
"description": "Base64 file contents, instead of file_url."
}
}
}We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 22, 2026.