Distill MCP Server

Let Claude, Cursor, or any MCP client rename files for you — automatically, based on content.

This is the official Model Context Protocol bridge for Distill — a native macOS app that uses AI (Claude, OpenAI, Gemini, Ollama, or Apple Intelligence) to read your files and rename them with descriptive, consistent names. No more IMG_4521.jpg or Scan_003.pdf — get 250401 Invoice Telekom.pdf instead.

You need the Distill app installed and running. This MCP server is a stdio→HTTP bridge — the actual file analysis happens in the app. Get Distill at store.juergenkoller.software/apps/distill.

What you can do

"Claude, look at the 47 PDFs in ~/Downloads, rename them based on content, but show me the suggestions first."

The MCP server exposes 9 tools:

Tool	What it does
`rename_files`	Analyze file content with AI and rename with descriptive names
`suggest_names`	Suggest new names without renaming (preview mode)
`revert_rename`	Undo a previous rename
`get_rename_history`	Show the rename history (timestamps, before/after, AI provider used)
`watch_folder`	Add a folder to auto-monitoring (new files get renamed in the background)
`app_status`	Current state, AI provider, credits left, watched folders
`set_provider`	Switch AI provider — Claude, OpenAI GPT-4o, Gemini, Ollama (local), Apple Intelligence
`get_rules`	Show naming rules (date format, casing, categories, custom templates)
`set_rules`	Update naming rules

Distill reads PDFs, images (via OCR), Office documents, emails, HTML, RTF, and media files. It extracts date, category, description, sender, and amounts — then builds filenames using your configurable rules.

Installation

Prerequisites

macOS 14 (Sonoma) or later
Distill app installed and running — get it here (free, pay-per-use credits start at €1.99 / 100 renames)
Swift 5.9+ (Xcode 15+) if building from source

Build from source

bash

git clone https://github.com/juergenkoller-software/distill-mcp.git
cd distill-mcp
swift build -c release
# Binary: .build/release/DistillMCP

Pre-built binary

Grab the latest DistillMCP from Releases.

Configuration

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

config.json

{
  "mcpServers": {
    "distill": {
      "command": "/path/to/DistillMCP",
      "env": {
        "DISTILL_PORT": "22200",
        "DISTILL_TOKEN": "your-token-here"
      }
    }
  }
}

Find DISTILL_TOKEN in Distill → Settings → API & MCP.

Claude Code

Terminal

claude mcp add distill /path/to/DistillMCP \
  --env DISTILL_PORT=22200 \
  --env DISTILL_TOKEN=your-token-here

Cursor / other MCP clients

Same pattern — DistillMCP is a stdio MCP server, configured via the two env vars above.

How it works

Code

┌────────────────┐  JSON-RPC stdio   ┌────────────────┐  HTTP+Bearer   ┌────────────────┐
│  Claude/Cursor │ ───────────────►  │  DistillMCP    │ ─────────────► │  Distill.app   │
│  (MCP client)  │ ◄───────────────  │   (this repo)  │ ◄───────────── │  (port 22200)  │
└────────────────┘                   └────────────────┘                └────────────────┘

The bridge reads JSON-RPC 2.0 requests from stdin, forwards them to Distill's local HTTP server at 127.0.0.1:22200/mcp, and writes responses back to stdout. All AI calls (Claude/OpenAI/Gemini/Ollama/Apple Intelligence), file reading (OCR, PDF extraction, Office parsing), credit tracking, and naming logic happen inside the Distill app.

This split lets the wire format stay open-source (audit it, sandbox it, run it through any MCP runtime) while the heavy lifting stays in the app.

Environment variables

Variable	Default	Description
`DISTILL_PORT`	`22200`	Port of Distill's local HTTP server
`DISTILL_TOKEN`	(none)	Bearer token from Distill Settings (required)

Errors and trace logs go to stderr so they don't pollute the JSON-RPC stdout channel.

About Distill

Distill is an AI file manager for macOS that automatically renames files based on content analysis. Highlights:

No subscription — free app, pay-per-use credits (100 free / 100 for €1.99 / 400 for €4.99 / 1,000 for €8.99)
Five AI providers — Claude, OpenAI GPT-4o, Google Gemini, Ollama (local, no API needed), Apple Intelligence
Reads everything — PDFs, images (OCR), Office docs, emails, HTML, RTF, media files
Local-only mode with Ollama — nothing leaves your Mac
Configurable NameBuilder — date formats, casing, categories, custom templates
Folder monitoring — drop files in, get them renamed automatically
REST API + Swagger UI — for non-MCP automation
MCP server (this repo) — for Claude/AI agents
Available on Mac App Store + direct download

→ Get Distill at store.juergenkoller.software

License

MIT — see LICENSE. Bridge is open source; the Distill app is commercial (free-to-try with pay-per-use credits).

Issues & support

Bridge bugs: open an issue
App support: support@juergenkoller.software

Built by Juergen Koller Software GmbH.