Docmost MCP Server

A Model Context Protocol (MCP) server that connects AI assistants (like Claude, Cursor, etc.) to standard Docmost instances.

Installation

# npm
npm install -g docmost-mcp

# or run directly with npx
npx docmost-mcp

Features

• Works with any standard Docmost installation (no modifications needed)
• Full support for spaces, pages, comments, search, and more
• Cookie-based JWT authentication
• Docker deployment with auto-restart

Available Tools

Space Tools

| Tool | Description | |------|-------------| | docmost_list_spaces | List all spaces in the workspace | | docmost_get_space | Get details of a specific space | | docmost_create_space | Create a new space | | docmost_update_space | Update an existing space | | docmost_delete_space | Delete a space |

Page Tools

| Tool | Description | |------|-------------| | docmost_list_pages | List pages in a space | | docmost_get_page | Get page details and content | | docmost_create_page | Create a new page | | docmost_update_page | Update page metadata | | docmost_update_page_markdown | Update page content with markdown | | docmost_delete_page | Delete a page | | docmost_get_recent_pages | Get recently updated pages | | docmost_get_page_history | Get page edit history | | docmost_move_page | Move page to new position | | docmost_move_page_to_parent | Move page to different parent or root | | docmost_bulk_create_pages | Create multiple pages at once | | docmost_import_page | Import page from markdown/HTML | | docmost_export_page | Export page as HTML/Markdown/PDF |

Attachment & Embedding Tools

| Tool | Description | |------|-------------| | docmost_upload_attachment | Upload image/file to a page | | docmost_get_attachment_info | Get attachment metadata by ID | | docmost_embed_image | Upload and insert image in one operation | | docmost_embed_from_cdn | Embed image from CDN URL (link or transfer modes) |

Diagram Tools

| Tool | Description | |------|-------------| | docmost_insert_drawio_block | Insert native Draw.io diagram from base64 SVG | | docmost_insert_drawio_from_cdn | Insert Draw.io diagram from CDN URL | | docmost_insert_excalidraw_block | Insert native Excalidraw diagram from base64 SVG | | docmost_insert_excalidraw_from_cdn | Insert Excalidraw diagram from CDN URL |

Search Tools

| Tool | Description | |------|-------------| | docmost_search | Search for pages in a space | | docmost_search_suggest | Get search suggestions |

Comment Tools

| Tool | Description | |------|-------------| | docmost_list_comments | List comments on a page | | docmost_create_comment | Create a comment | | docmost_delete_comment | Delete a comment |

Workspace Tools

| Tool | Description | |------|-------------| | docmost_get_workspace | Get workspace info | | docmost_list_workspace_members | List workspace members | | docmost_get_current_user | Get current user info |

Page Composition Order

Docmost pages are TipTap documents with two flavours of content:

• Markdown-equivalent nodes — headings, paragraphs, lists, tables, code blocks, blockquotes. These round-trip cleanly through docmost_update_page_markdown.
• Native TipTap nodes — drawio, excalidraw, and image attachments. These have no markdown equivalent and only exist as structured nodes in the document tree.

When composing a page, build it in this order:

1. Markdown body first with docmost_update_page_markdown (operation: replace).
2. Diagram blocks last via docmost_insert_drawio_block, docmost_insert_drawio_from_cdn, docmost_insert_excalidraw_block, or docmost_insert_excalidraw_from_cdn.
3. Image embeds anywhere with docmost_embed_from_cdn — these correctly append/prepend to existing content of either kind.

Avoid calling docmost_update_page_markdown(replace) after you've inserted diagram blocks. The replace operation rewrites the page from markdown, and since native nodes have no markdown form, they're dropped. Use docmost_update_page_markdown(append) or docmost_update_page_markdown(prepend) to add prose around existing diagram blocks without losing them.

For documents that mix prose and diagrams, the cleanest workflow is:

1. Build out the markdown body (replace).
2. prepend any header/title images.
3. append diagram blocks at the end, in order.
4. Use append operations on update_page_markdown to add explanatory prose between or after diagram blocks.

Setup

1. Authentication

You can authenticate using either email/password or an auth token:

Option A: Email/Password (recommended)

DOCMOST_URL=https://your-docmost-instance.com
[email protected]
DOCMOST_PASSWORD=your-password

Option B: Auth Token

1. Log into your Docmost instance in a browser
2. Open Developer Tools (F12) → Application → Cookies
3. Copy the authToken cookie value

DOCMOST_URL=https://your-docmost-instance.com
DOCMOST_AUTH_TOKEN=your_auth_token_here

2. Create Environment File

cp .env.example .env
# Edit .env with your credentials

3. Run with Docker

docker compose up -d

Usage with Claude Code

Add to your Claude Code MCP configuration (~/.claude.json or settings):

{
  "mcpServers": {
    "docmost": {
      "command": "node",
      "args": ["/path/to/docmost-mcp/dist/index.js"],
      "env": {
        "DOCMOST_URL": "https://your-docmost-instance.com",
        "DOCMOST_EMAIL": "[email protected]",
        "DOCMOST_PASSWORD": "your-password"
      }
    }
  }
}

Or with Docker:

{
  "mcpServers": {
    "docmost": {
      "command": "docker",
      "args": ["exec", "-i", "docmost-mcp", "node", "dist/index.js"],
      "env": {}
    }
  }
}

Local Development

# Install dependencies
pnpm install

# Build
pnpm run build

# Run in development mode
pnpm run dev

Important Notes

• The auth token expires after 30 days. You'll need to get a new one periodically.
• This uses internal Docmost API endpoints which may change with updates.
• Some operations require specific permissions in Docmost.

License

MIT