Skip to main content
Glama

RapidOCR MCP Server

High-performance OCR MCP Server.

Features

  • Multi-mode: MCP stdio, FastAPI HTTP, streamable-http

  • Multi-input: Path, Base64, URL, File upload

  • Batch OCR: Process multiple images in one request

  • Output formats: Plain, JSON, Markdown, Structured

  • Image preprocessing: Auto-enhance, rotate, binarize

  • Security: Path whitelist, API key, CORS, audit logging

  • Monitoring: Prometheus metrics, OpenTelemetry tracing

  • Production-ready: Docker, CI/CD, tests

Related MCP server: mcp-ocrspace

Quick Start

Global Installation

# Install globally with uvx (recommended)
uvx rapidocr-mcp

# Or install globally with pip
pip install rapidocr-mcp

Local Development

# Install dependencies
uv sync

# Run MCP server (stdio mode)
uv run rapidocr-mcp

# Run FastAPI server
uv run rapidocr-mcp --mode fastapi --port 8080

Configuration

Set environment variables with RAPIDOCR_ prefix:

export RAPIDOCR_LANG=ch
export RAPIDOCR_LOG_LEVEL=INFO
export RAPIDOCR_API_KEY=your-key

MCP Configuration

Available Tools

Tool

Description

ocr_by_path

OCR for local image file by path

ocr_by_content

OCR for Base64 encoded image

ocr_by_url

OCR for image from HTTP/HTTPS URL

ocr_batch

Batch OCR for multiple images

Claude Desktop

Add to your Claude Desktop config file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "rapidocr": {
      "command": "uvx",
      "args": ["rapidocr-mcp"]
    }
  }
}

Or with local installation:

{
  "mcpServers": {
    "rapidocr": {
      "command": "uv",
      "args": ["--directory", "/path/to/rapidocr-mcp", "run", "rapidocr-mcp"]
    }
  }
}

Other MCP Clients

For other MCP clients that support stdio mode, configure the command as:

uvx rapidocr-mcp
# or
rapidocr-mcp

Docker

docker-compose -f docker/docker-compose.yml up

API Endpoints

  • GET /health - Health check

  • POST /ocr/path - OCR by file path

  • POST /ocr/base64 - OCR by base64

  • POST /ocr/url - OCR by URL

  • POST /ocr/upload - OCR by file upload

  • GET /metrics - Prometheus metrics

Sponsoring

If you find this project helpful, consider buying me a coffee!

License

MIT

Available Tools

4 tools
ocr_batchA

批量 OCR - 对多张图像执行 OCR 识别。

Args:
    image_paths: 图像文件路径列表
    output_format: 输出格式 (plain/json/markdown/structured)
    auto_enhance: 自动增强图像对比度和锐度
    rotate: 自动旋转图像(基于 EXIF)
    binarize: 二值化图像

Returns:
    所有图像的 OCR 识别结果

Inputs: image_paths (required), rotate, binarize, auto_enhance, +1 more.

ocr_by_contentA

对 Base64 编码的图像执行 OCR 识别。

Args:
    image_base64: Base64 编码的图像(支持 data:image/... 前缀)
    output_format: 输出格式 (plain/json/markdown/structured)
    auto_enhance: 自动增强图像对比度和锐度
    rotate: 自动旋转图像(基于 EXIF)
    binarize: 二值化图像

Returns:
    OCR 识别结果

Inputs: image_base64 (required), rotate, binarize, auto_enhance, +1 more.

ocr_by_pathA

对本地图像文件执行 OCR 识别。

Args:
    image_path: 图像文件的绝对路径或相对路径
    output_format: 输出格式 (plain/json/markdown/structured)
    auto_enhance: 自动增强图像对比度和锐度
    rotate: 自动旋转图像(基于 EXIF)
    binarize: 二值化图像

Returns:
    OCR 识别结果

Inputs: image_path (required), rotate, binarize, auto_enhance, +1 more.

ocr_by_urlA

对 URL 图像执行 OCR 识别。

Args:
    image_url: 图像的 HTTP/HTTPS URL
    output_format: 输出格式 (plain/json/markdown/structured)
    use_cache: 是否使用 URL 缓存
    auto_enhance: 自动增强图像对比度和锐度
    rotate: 自动旋转图像(基于 EXIF)
    binarize: 二值化图像

Returns:
    OCR 识别结果

Inputs: image_url (required), rotate, binarize, use_cache, +2 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.2/5.0

    Scored across 4 tools

    Disambiguation5/5

    Each tool targets a distinct input source: local file path, base64 content, URL, and batch of paths. The descriptions clearly differentiate the tools, and the batch tool's multi-image purpose avoids confusion with single-image processing.

    Naming Consistency4/5

    Three tools follow the 'ocr_by_<source>' pattern clearly, but 'ocr_batch' deviates by omitting 'by'. This is a minor inconsistency that does not harm readability.

    Tool Count5/5

    Four tools are well-scoped for an OCR server, covering the primary input methods (file, content, URL, batch) without unnecessary redundancy. The count feels complete and focused.

    Completeness5/5

    The tool surface covers the core OCR workflows for local files, base64 data, URLs, and batch processing. There are no obvious gaps for a typical OCR use case, making the set comprehensive.

    Maintenance

    ActivityInactive
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers