# MarkIndex MCP server

> Enables AI agents to perform structure-first, vectorless retrieval and keyword search over large Markdown libraries using section trees and BM25.

- Repository: https://github.com/jackjin1997/markindex
- Page: https://hostedmcp.org/servers/jackjin1997/markindex
- Transport: stdio
- Stars: 1
- License: MIT
- Latest version: 0.1.2

## Install

### Claude Code

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

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

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

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

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

### VS Code CLI

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

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

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

### Codex

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

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

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

### Gemini CLI

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

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

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

## Tools (5)

- `browse_documents`
- `get_document`
- `get_document_structure`
- `get_node_content`
- `search_documents`

## FAQ

### Does MarkIndex require a vector database?

No, it is a vectorless, structure-first retrieval system that uses section trees and BM25 indexing stored in a SQLite file.

### How does the search handle Chinese and English text?

It uses BM25 over Intl.Segmenter word segmentation for Chinese, English, and mixed text, supplemented by a trigram substring index.
