# filegoblin MCP server

> Converts PDFs, Office documents, images, audio, and websites into clean, LLM-ready Markdown with layout-aware semantic chunking.

- Repository: https://github.com/funkyfunc/filegoblin
- Page: https://hostedmcp.org/servers/funkyfunc/filegoblin
- Transport: stdio
- Stars: 1
- License: MIT

## Install

### Claude Code

```sh
claude mcp add filegoblin -- npx -y filegoblin
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "filegoblin": {
      "command": "npx",
      "args": [
        "-y",
        "filegoblin"
      ]
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "filegoblin": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "filegoblin"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"filegoblin","type":"stdio","command":"npx","args":["-y","filegoblin"]}'
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "filegoblin": {
      "command": "npx",
      "args": [
        "-y",
        "filegoblin"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add filegoblin -- npx -y filegoblin
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.filegoblin]
command = "npx"
args = ["-y", "filegoblin"]
```

### Gemini CLI

```sh
gemini mcp add -s user filegoblin npx -- -y filegoblin
```

### Gemini settings.json (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "filegoblin": {
      "command": "npx",
      "args": [
        "-y",
        "filegoblin"
      ]
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `FILEGOBLIN_ALLOW_LOCAL_URLS` | no | no |  |

## FAQ

### Does filegoblin process files locally?

Yes, all parsing and PII scrubbing happen entirely on your machine without using cloud APIs.

### What file formats are supported?

It supports PDF, Word, Excel, PowerPoint, email, EPUB, HTML, Jupyter, Code, and more.

### Can it handle scanned documents?

Yes, it includes OCR capabilities for scanned PDFs and images, with specialized vision models for layout and reading order.

### How does it handle large documents for LLMs?

It features layout-aware semantic chunking that protects tables and code blocks and replicates table headers across splits.