PDF To Text

Name: PDF To Text
Author: Jim-Panse

本地服务

github.com/Jim-Panse生产力

代码仓库

关于 PDF To Text

在本地提取 PDF 文本，支持 macOS PDFKit 与 Vision OCR 作为备用方案，并可选择性使用 AI 重试。

配置

本地服务的配置在各客户端中通用，复制后粘贴到客户端的 MCP 配置文件即可。

mcp.json · 通用配置

{
  "mcpServers": {
    "jim-panse-pdf-to-text": {
      "command": "npx",
      "args": [
        "-y",
        "@jim-panse/pdf-to-text-mcp"
      ],
      "env": {
        "PDF_TO_TEXT_MCP_ROOT": "<PDF_TO_TEXT_MCP_ROOT>",
        "OPENAI_API_KEY": "<OPENAI_API_KEY>",
        "OPENAI_MODEL": "<OPENAI_MODEL>",
        "AI_MAX_OUTPUT_TOKENS": "<AI_MAX_OUTPUT_TOKENS>",
        "MAX_PARALLEL_EXTRACTIONS": "<MAX_PARALLEL_EXTRACTIONS>",
        "EXTRACTION_QUEUE_CAPACITY": "<EXTRACTION_QUEUE_CAPACITY>",
        "EXTRACTION_QUEUE_WAIT_MS": "<EXTRACTION_QUEUE_WAIT_MS>",
        "WRITE_TEXT": "<WRITE_TEXT>"
      }
    }
  }
}

启动命令npx -y @jim-panse/pdf-to-text-mcp

需要配置的环境变量

PDF_TO_TEXT_MCP_ROOTRequired existing directory that defines the root-relative filesystem namespace for PDF input and derived text output.

OPENAI_API_KEY密钥Optional secret. Enables one AI extraction retry when local extraction is inadequate; omit for local-only operation.

OPENAI_MODELOptional OpenAI model used only for AI retry. Defaults to gpt-5.4-mini.

AI_MAX_OUTPUT_TOKENS密钥Optional cap for all AI-generated output tokens, including non-visible reasoning tokens. Integer minimum 800; defaults to 32000.

MAX_PARALLEL_EXTRACTIONSInteger concurrency limit from 1 to 32. Defaults to min(os.availableParallelism(), 4).

EXTRACTION_QUEUE_CAPACITYInteger maximum waiting requests from 0 to 100; zero disables waiting.

EXTRACTION_QUEUE_WAIT_MSMaximum queue wait in milliseconds, integer 1 through 2147483647. Does not limit active work.

WRITE_TEXTOptionally save successful extraction text beside the source PDF.

本地服务需要在你的电脑上安装运行环境（如 Node.js 18+）。各客户端的配置文件位置不同，请参考客户端文档。

安装包

npm@jim-panse/pdf-to-text-mcp0.7.0stdio