Skip to main content
Glama
h4sht

@h4sht/bookstack-mcp

by h4sht
OverviewSchemaSourceRelated ServersDiscussionsAdmin
TypeScript

@h4sht/bookstack-mcp

🚀 Claude Code

claude mcp add bookstack -- npx github:h4sht/bookstack-mcp

Claude Code will prompt you for:

  • BOOKSTACK_BASE_URL — your BookStack API URL (e.g. https://wiki.example.com/api)

  • BOOKSTACK_API_TOKEN — your API token (token_id:token_secret)

Or pass them directly:

claude mcp add bookstack \
  --env BOOKSTACK_BASE_URL=https://wiki.example.com/api \
  --env BOOKSTACK_API_TOKEN=token_id:token_secret \
  -- npx github:h4sht/bookstack-mcp

Then use /mcp inside Claude Code to verify it's connected ✔

Related MCP server: BookStack MCP Server

🛠️ What you get (25 tools)

Category

Tools

📚 Books (6)

list, get, create, update, delete, export

📑 Chapters (6)

list, get, create, update, delete, export

📄 Pages (6)

list, get, create, update, delete, export

📚 Shelves (5)

list, get, create, update, delete

🔍 Search (1)

search across all content

⚙️ System (1)

instance info

All export_* tools support markdown, html, pdf, plaintext.

✍️ Markdown-powered

BookStack stores pages as HTML but accepts Markdown input through its API. This means Claude can create and edit wiki pages in its native format:

"Create a page about async Python in book 3"

→ bookstack_create_page({
    book_id: 3,
    name: "Async Python Guide",
    markdown: "# Async Python\n\n`asyncio` is Python's standard library..."
  })

Pages are also exported as Markdown, so Claude can read, analyze, and improve existing content.

📦 Setup for other AI assistants

OpenAI Codex

codex mcp add bookstack \
  --env BOOKSTACK_BASE_URL=https://wiki.example.com/api \
  --env BOOKSTACK_API_TOKEN=id:secret \
  -- npx github:h4sht/bookstack-mcp

Or add this to ~/.codex/config.toml:

[mcp_servers.bookstack]
command = "npx"
args = ["github:h4sht/bookstack-mcp"]

[mcp_servers.bookstack.env]
BOOKSTACK_BASE_URL = "https://wiki.example.com/api"
BOOKSTACK_API_TOKEN = "id:secret"

Verify with: codex mcp list

OpenCode

opencode mcp add
# Select "local", then enter: npx github:h4sht/bookstack-mcp

Or add to opencode.json:

{
  "mcp": {
    "bookstack": {
      "type": "local",
      "command": ["npx", "github:h4sht/bookstack-mcp"],
      "enabled": true,
      "environment": {
        "BOOKSTACK_BASE_URL": "https://wiki.example.com/api",
        "BOOKSTACK_API_TOKEN": "id:secret"
      }
    }
  }
}

Cline (VS Code)

Add to Cline MCP settings file:

{
  "mcpServers": {
    "bookstack": {
      "command": "npx",
      "args": ["github:h4sht/bookstack-mcp"],
      "env": {
        "BOOKSTACK_BASE_URL": "https://wiki.example.com/api",
        "BOOKSTACK_API_TOKEN": "id:secret"
      }
    }
  }
}

Any MCP-compatible assistant (stdio)

The server command is always the same:

npx github:h4sht/bookstack-mcp

Set BOOKSTACK_BASE_URL and BOOKSTACK_API_TOKEN as environment variables.

🔧 Config reference

Variable

Required

Description

BOOKSTACK_BASE_URL

Yes

Full URL including /api

BOOKSTACK_API_TOKEN

Yes

Format: token_id:token_secret

BOOKSTACK_TIMEOUT

No

Request timeout ms (default: 30000)

🔒 Security

  • 4 auditable files — read the whole source in 10 minutes

  • One dependency — only @modelcontextprotocol/sdk (official Anthropic)

  • Native fetch() — no HTTP libraries, no hidden requests

  • Stdio only — no open ports, no network surface

  • Zero telemetry — no analytics, no tracking, no calls home

  • Stderr logging — never writes to stdout (MCP protocol integrity)

🧪 Tests

npm test   # 32 tests, all passing ✅

📁 Structure

@h4sht/bookstack-mcp/
├── src/
│   ├── index.ts      # MCP server, stdio transport, config
│   ├── client.ts     # BookStack REST client (native fetch)
│   ├── tools.ts      # 25 MCP tool definitions + handlers
│   └── types.ts      # TypeScript types
├── dist/             # Compiled JS (ready to run)
├── tests/            # 32 unit + integration tests
├── package.json
└── README.md

📝 License

MIT © h4sht

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityStale
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers