Distill Mcp

io.github.lakshgk由 GitHub 账号 lakshgk 发布v0.1.1更新于 Oct 24, 2026

Convert PDF, DOCX, XLSX, PPTX, HTML, audio to token-efficient Markdown for any MCP client.

已验证STDIO仅桌面文件与存储知识与记忆

概览

AI 生成的概览

将本地 PDF、DOCX、XLSX、PPTX、HTML 和音频等文档转换为节省 token 的 Markdown，供 MCP 客户端使用。

可以做什么
Distill Mcp 提供 convert_and_save 工具，在模型读取之前把本地文件转换为干净的 Markdown，相比原始文档文本通常可减少 40% 到 80% 的 token。它支持 Word、Excel、PowerPoint、原生文本 PDF、HTML、EPUB、JSON、SQL 和 WSDL 文件；完整模式还支持通过 OCR 处理扫描版 PDF 以及音频格式。转换结果会被缓存，工具会报告是否覆盖了之前的缓存版本以及任何警告。
适用场景
适合助手需要读取或推理本地文档且上下文预算有限的场景，例如总结报告、表格或演示文稿。也适合客户端被配置为先转换文件、而不是通过文件系统工具直接读取原始文件的流程。
运行要求
通过 stdio 在本地运行，使用 npx -y distill-mcp-server 启动，因此需要 Node.js。轻量模式还需要 Python 和 pip，通过环境变量 DISTILL_MCP_CONFIG 配置 python_path。完整模式需要 Docker 并在本地运行 Distill 服务，通过 DISTILL_MCP_CONFIG 配置 distill_url。未声明需要账号或 API 密钥。
安装前请注意
该服务器按路径读取本地文件并写入转换后的 Markdown，可能覆盖之前缓存的转换结果，工具会通过 overwritten 标志报告。完整模式会把文档内容发送到本地 Docker 服务而非外部服务，但该服务必须正在运行且可访问。README 声明所有处理均在本地进行，数据不会离开本机。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Distill Mcp，将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Desktop only，通过 STDIO。 STDIO 服务会启动本地进程，因此需要 SourceWeft 桌面宿主。

其他 MCP 客户端

参照 仓库 中的启动说明。

README

distill-mcp

[npm version] [license] [MCP Registry]

An MCP server that connects any MCP-compatible client to Distill, converting local documents to clean, token-efficient Markdown before the LLM reads them. Typical token reduction is 40–80% compared to raw document text, letting the model fit more content into its context window and reason over it faster.

Works with Claude Desktop, Claude Code, Cursor, Windsurf, and any other MCP client that supports tool calling.

Two Modes

LightweightFull (Docker)
Requires DockerNoYes
DOCX, XLSX, PPTX, native PDF, HTMLYesYes
Scanned PDF (OCR)NoYes
Audio (MP3, WAV, etc.)NoYes
Quality scoreNoYes
Install time~2 min~10 min

Lightweight needs only Python and pip. Full needs Docker with the Distill service running. See the setup guides for details.

Quick install

No global install required:

bash
npx -y distill-mcp-server

Then configure Claude Desktop — see Lightweight setup or Full setup for step-by-step instructions.

Configuration

Add the server to your claude_desktop_config.json.

Lightweight mode — macOS / Linux:

json
{  "mcpServers": {    "distill-mcp": {      "command": "npx",      "args": ["-y", "distill-mcp-server"],      "env": {        "DISTILL_MCP_CONFIG": "{\"mode\":\"lightweight\",\"python_path\":\"python3\"}"      }    }  }}

Lightweight mode — Windows:

json
{  "mcpServers": {    "distill-mcp": {      "command": "npx",      "args": ["-y", "distill-mcp-server"],      "env": {        "DISTILL_MCP_CONFIG": "{\"mode\":\"lightweight\",\"python_path\":\"py\"}"      }    }  }}

Full mode — all platforms:

json
{  "mcpServers": {    "distill-mcp": {      "command": "npx",      "args": ["-y", "distill-mcp-server"],      "env": {        "DISTILL_MCP_CONFIG": "{\"mode\":\"full\",\"distill_url\":\"http://localhost:7860\"}"      }    }  }}

Full config key reference is in the setup guides: Lightweight | Full

CLAUDE.md snippet

Paste this into your project's CLAUDE.md so Claude knows how to use the tool automatically. Copy the block below as-is:

markdown
## Document Conversion — distill-mcp
When the user references a local file path (e.g. a PDF, DOCX, PPTX, XLSX,HTML, or audio file), ALWAYS call the `convert_and_save` tool before readingor reasoning about the document. Do not read the original file directly viafilesystem tools — use only the Markdown returned by `convert_and_save` asthe document content.
### Rules
1. Call `convert_and_save` with the absolute file path before doing anything   else with the document.2. Use ONLY the Markdown output from `convert_and_save` as the document   content. Never read the original file with filesystem tools.3. If the response includes `"overwritten": true`, tell the user that a   previous cached version was replaced before proceeding.4. If the response includes any `warnings`, surface them to the user before   proceeding with the document content.5. If `convert_and_save` returns an unsupported format error, tell the user   which formats are supported and suggest switching modes if applicable.
### Supported formats
| Category | Lightweight | Full (Docker) ||---|---|---|| Word | .docx, .doc, .odt | .docx, .doc, .odt || Excel | .xlsx, .xlsm, .csv | .xlsx, .xlsm, .csv || PowerPoint | .pptx, .ppt | .pptx, .ppt || PDF | .pdf (native text) | .pdf (native + scanned OCR) || HTML | .html, .htm | .html, .htm || Audio | — | .mp3, .wav, .m4a, .flac, .ogg || Other | — | .epub, .json, .sql, .wsdl, .wsd |
### Usage
Say "convert using distill" followed by the file path:> Convert using distill C:\Users\me\Documents\report.pdf to markdown

The snippet is also available in docs/CLAUDE.md-snippet.md.

Supported formats

CategoryExtensionsLightweightFull
Microsoft Word.docx, .doc, .odtYesYes
Microsoft Excel.xlsx, .xlsm, .csvYesYes
Microsoft PowerPoint.pptx, .pptYesYes
PDF (native text).pdfYesYes
PDF (scanned/OCR).pdfNoYes
HTML.html, .htmYesYes
Audio.mp3, .wav, .m4a, .flac, .oggNoYes
EPUB.epubNoYes
JSON.jsonNoYes
SQL.sqlNoYes
WSDL.wsdl, .wsdNoYes

Privacy

All processing happens locally. In lightweight mode, documents are converted by the distill-core Python library on your machine — no data leaves your computer. In full mode, documents are sent to the Distill Docker service running locally on your machine — no data is sent to external services.

Built on

Distill — document-to-Markdown conversion engine.

License

MIT

来源：README.md，提交 da01804

工具

0
SourceWeft 尚未在沙箱中启动过该软件包，因此其工具未知。

版本历史

1
  1. v0.1.1最新Sep 16, 2026

评分

已满足适用检查项的 50%

该评分衡量此收录条目的完整程度，依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价，也不表示服务器安全或当前可用：可连通性是带日期的检查结果，评分也不包含安装量或用户评价。

  • 近 30 天内可连通25 分

    请保持端点能够响应 MCP 握手；SourceWeft 每 30 天探测一次。

  • 已声明认证方式10 分

    请在注册中心条目中声明凭据请求头，或保持端点能够响应探测，以便判定其认证方式。

  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list，或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分