notion-mcp-server

English | 日本語

MCP (Model Context Protocol) server for Notion API. Enables AI assistants to interact with Notion pages, databases, and blocks.

API Version: 2025-09-03 (latest)

Why this repository?

Why I Built This

I wanted to use AI agents to process tasks from my Notion database with specific conditions:

"Get tasks where Status = 'Not Started' AND Assignee = 'Alice', sorted by Priority"

Here's what I found with existing options:

The gap I wanted to fill:

• Full database property filtering (AND/OR, select, checkbox, date, etc.)

• Optimized response sizes for LLM token efficiency

• No plan restrictions

This repository provides property filtering with fields parameter for 90% token reduction.

Related MCP server: Notion MCP Server

Quick Start

1. Get a Notion Token

1. Go to Notion Integrations

2. Click "New integration"

3. Give it a name and select the workspace

4. Copy the "Internal Integration Token" (starts with ntn_)

5. Share the pages/databases you want to access with your integration

2. Configure Your AI Client

Claude Desktop

Add to your configuration (~/.config/claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "notion": {
      "command": "npx",
      "args": ["-y", "@atikk-co-jp/notion-mcp-server"],
      "env": {
        "NOTION_TOKEN": "ntn_xxxxxxxxxxxx"
      }
    }
  }
}

Claude Code

Add to your .mcp.json:

{
  "mcpServers": {
    "notion": {
      "command": "npx",
      "args": ["-y", "@atikk-co-jp/notion-mcp-server"],
      "env": {
        "NOTION_TOKEN": "ntn_xxxxxxxxxxxx"
      }
    }
  }
}

That's it! Restart your AI client and start using Notion.

Features

• Page Operations: Create, retrieve, update, and move Notion pages

• Database Operations: Create, retrieve, update, and query databases with filters and sorts

• Block Operations: Retrieve, update, delete, and append blocks

• Search: Search across pages and databases

• Comments: Create and list comments

• Users: List users and retrieve user info

• Token-Efficient Output: Markdown/simple format reduces token usage by ~96%

• Markdown Input: Create and append content using Markdown (80% fewer input tokens)

API Coverage

⭐ = Markdown input supported (reduces input tokens by ~80%)

📤 = Minimal response (id/url only) - reduces output tokens by ~90%

Available Tools

retrieve-page

Retrieve a Notion page by its ID.

Parameters:

• page_id (required): The ID of the page to retrieve

• format (optional): Output format - "simple" (default) or "json"

  • simple: Returns simplified property values with reduced token usage

  • json: Returns raw Notion API response

• include_content (optional): Include page content as markdown (default: true)

{
  "page_id": "page-uuid-here",
  "format": "simple",
  "include_content": true
}

create-page

Create a new page. Supports two parent types:

• Child page: Use parent.page_id to create a page under an existing page

• Database entry: Use parent.data_source_id to create a page in a database

Create a child page:

{
  "parent": { "page_id": "parent-page-uuid-here" },
  "properties": {
    "title": {
      "title": [{ "text": { "content": "Child Page Title" } }]
    }
  }
}

Create a database entry:

{
  "parent": { "data_source_id": "data-source-uuid-here" },
  "properties": {
    "Name": {
      "title": [{ "text": { "content": "New Page Title" } }]
    },
    "Status": {
      "status": { "name": "In Progress" }
    }
  }
}

create-page-simple ⭐

Create a new page using Markdown. ~80% fewer input tokens compared to create-page.

Supports two parent types:

• Child page: Use parent.page_id to create a page under an existing page

• Database entry: Use parent.data_source_id to create a page in a database

Parameters:

• parent (required): Either { page_id } or { data_source_id }

• title (required): Page title as a simple string

• content (optional): Page content in Markdown

• properties (optional): Additional Notion properties

• icon (optional): Emoji icon (e.g., "🐛")

Supported Markdown:

• Headings: # ## ### (#### and beyond → heading_3)

• Lists: - or * (bulleted), 1. (numbered)

• Checkboxes: - [ ] / - [x]

• Code blocks: ``` with language

• Quotes: >

• Dividers: ---

• Images: ![alt](url)

• Tables: | col1 | col2 | with header separator |---|---|

• Inline: **bold**, *italic*, ~~strike~~, `code`, [link](url)

Extended Markdown (bidirectional):

• Toggle: <details><summary>title</summary>content</details>

• Callout: > [!NOTE], > [!WARNING], > [!TIP], > [!IMPORTANT], > [!CAUTION]

• Equation: $$E = mc^2$$ (inline/block)

• Underline: <u>text</u> or ++text++

• Color: {color:red}text{/color}, {bg:yellow}text{/bg}

• Bookmark: [bookmark](url) or [bookmark:caption](url)

• Columns: :::columns / :::column / :::

• Media: @[embed](url), @[video](url), @[audio](url), @[file](url), @[pdf](url)

• Table of contents: [TOC]

Create a child page:

{
  "parent": { "page_id": "parent-page-uuid-here" },
  "title": "Meeting Notes",
  "content": "## Agenda\n\n1. Review progress\n2. Next steps"
}

Create a database entry:

{
  "parent": { "data_source_id": "data-source-uuid-here" },
  "title": "Bug Report",
  "content": "## Steps to Reproduce\n\n1. Login\n2. Open settings\n\n## Expected Behavior\n\nShould display correctly",
  "properties": {
    "Status": { "status": { "name": "Open" } }
  },
  "icon": "🐛"
}

Token Comparison:

update-page

Update a page's properties, icon, cover, archive status, or lock status.

Parameters:

• page_id (required): The ID of the page to update

• properties (optional): Properties to update

• icon (optional): Icon (set to null to remove)

• cover (optional): Cover image (set to null to remove)

• archived (optional): Set to true to archive

• is_locked (optional): Lock the page to prevent edits in the UI

{
  "page_id": "page-uuid-here",
  "properties": {
    "Status": {
      "status": { "name": "Done" }
    }
  },
  "is_locked": true
}

query-data-source

Query a data source with optional filters and sorts.

Parameters:

• data_source_id (required): The ID of the data source to query

• filter (optional): Filter conditions as a JSON object

• sorts (optional): Sort conditions as an array

• start_cursor (optional): Cursor for pagination

• page_size (optional): Number of results to return (1-100)

• format (optional): Output format - "simple" (default) or "json"

  • simple: Returns simplified property values with reduced token usage

  • json: Returns raw Notion API response

{
  "data_source_id": "data-source-uuid-here",
  "filter": {
    "property": "Status",
    "status": { "equals": "In Progress" }
  },
  "sorts": [
    { "property": "Created", "direction": "descending" }
  ],
  "format": "simple"
}

create-database

Create a new database as a subpage of an existing page.

Parameters:

• parent_page_id (required): The ID of the parent page

• properties (required): Database schema with at least one title property

• title (optional): Database title as rich text array

• icon (optional): Icon for the database

• cover (optional): Cover image for the database

• is_inline (optional): If true, creates an inline database

{
  "parent_page_id": "parent-page-uuid",
  "properties": {
    "Name": { "title": {} },
    "Status": { "select": { "options": [{ "name": "Todo" }, { "name": "Done" }] } },
    "Priority": { "number": {} }
  },
  "title": [{ "type": "text", "text": { "content": "Task Database" } }]
}

update-database

Update an existing database container (title, description, icon, cover).

Note: For schema (properties/columns) updates, use update-data-source instead.

Parameters:

• database_id (required): The ID of the database to update

• title (optional): New title as rich text array

• description (optional): New description as rich text array

• icon (optional): Icon (set to null to remove)

• cover (optional): Cover image (set to null to remove)

• is_inline (optional): If true, creates an inline database

• archived (optional): Set to true to archive

• is_locked (optional): Lock the database to prevent edits in the UI

{
  "database_id": "database-uuid-here",
  "title": [{ "type": "text", "text": { "content": "New Title" } }],
  "is_locked": true
}

retrieve-data-source

Retrieve a data source schema by its ID.

Parameters:

• data_source_id (required): The ID of the data source

• format (optional): Output format - "simple" (default) or "json"

{
  "data_source_id": "data-source-uuid-here",
  "format": "simple"
}

update-data-source

Update a data source schema (properties/columns).

Parameters:

• data_source_id (required): The ID of the data source to update

• properties (optional): Properties to add, update, or delete (set to null)

{
  "data_source_id": "data-source-uuid-here",
  "properties": {
    "NewColumn": { "rich_text": {} },
    "OldColumn": null
  }
}

search

Search across all pages and data sources.

{
  "query": "search term",
  "filter": { "value": "page", "property": "object" }
}

Filter values: "page" or "data_source"

get-block-children

Get the child blocks of a page or block.

Parameters:

• block_id (required): The ID of the block or page to get children from

• start_cursor (optional): Cursor for pagination

• page_size (optional): Number of results to return (1-100)

• format (optional): Output format - "markdown" (default), "simple", or "json"

  • markdown: Returns human-readable markdown with significantly reduced token usage (~96% reduction)

  • simple: Returns ID + type + markdown content (lightweight, for deletion target selection)

  • json: Returns raw Notion API response

• fetch_nested (optional): When format="markdown", fetch nested children blocks recursively (default: false)

{
  "block_id": "page-or-block-uuid-here",
  "format": "markdown",
  "fetch_nested": true
}

Get block IDs for deletion:

{
  "block_id": "page-or-block-uuid-here",
  "format": "simple"
}

Returns:

{
  "blocks": [
    { "id": "abc123", "type": "heading_1", "content": "# Title" },
    { "id": "def456", "type": "paragraph", "content": "Some text" }
  ],
  "has_more": false
}

append-block-children

Append new blocks to a page or block.

{
  "block_id": "page-or-block-uuid-here",
  "children": [
    {
      "type": "paragraph",
      "paragraph": {
        "rich_text": [{ "text": { "content": "New paragraph" } }]
      }
    }
  ]
}

append-blocks-simple ⭐

Append blocks using Markdown. ~80% fewer output tokens compared to append-block-children.

Parameters:

• block_id (required): The page or block ID to append to

• content (required): Content in Markdown

• after (optional): Insert after this block ID

Same Markdown support as create-page-simple.

{
  "block_id": "page-or-block-uuid-here",
  "content": "# New Section\n\nThis is **important** content with a [link](https://example.com).\n\n- Item 1\n- Item 2\n\n```javascript\nconst x = 1;\n```"
}

Token Comparison:

replace-page-content ⭐

Replace all content of a page with new Markdown content. Automatically preserves child_database and child_page blocks.

Parameters:

• page_id (required): The page ID to update

• content (required): New content in Markdown

• dry_run (optional): Preview which blocks will be deleted without making changes (default: false)

⚠️ Warning: Blocks not representable in Extended Markdown (table_of_contents, synced_block, etc.) will be DELETED. Use dry_run: true to preview before executing.

Use when: You want to completely rewrite page content without finding individual block IDs.

Same Markdown support as create-page-simple.

{
  "page_id": "page-uuid-here",
  "content": "# New Page Title\n\nThis is the new content.\n\n## Section 1\n\n- Item 1\n- Item 2"
}

Preview deletions (dry run):

{
  "page_id": "page-uuid-here",
  "content": "# New content",
  "dry_run": true
}

find-and-replace-in-page ⭐

Find text in a page and replace it with new content. Supports regex patterns for advanced matching.

Parameters:

• page_id (required): The page ID to search in

• find (required): Text to find (string or regex pattern)

• replace (required): Replacement text (supports Markdown: **bold**, *italic*, etc.)

• use_regex (optional): If true, treat find as a regex pattern (default: false)

Use when: You want to update specific text without rewriting the entire page.

{
  "page_id": "page-uuid-here",
  "find": "old text",
  "replace": "**new text**"
}

With regex:

{
  "page_id": "page-uuid-here",
  "find": "item\\d+",
  "replace": "updated item",
  "use_regex": true
}

delete-blocks-batch

Delete multiple blocks by their IDs. Blocks are deleted sequentially to respect API rate limits (3 req/s).

Parameters:

• block_ids (required): Array of block IDs to delete (max 100)

Use when: You want to delete specific blocks. Use get-block-children with format="simple" to get block IDs first.

{
  "block_ids": ["block-uuid-1", "block-uuid-2", "block-uuid-3"]
}

Returns:

{
  "deleted_count": 3,
  "failed_count": 0,
  "deleted": ["block-uuid-1", "block-uuid-2", "block-uuid-3"]
}

clear-page-content

Delete all content from a page. By default, preserves child_database and child_page blocks.

Parameters:

• page_id (required): The page ID to clear

• preserve_types (optional): Block types to preserve (default: ["child_database", "child_page"]). Set to [] to delete all.

Use when: You want to delete all content from a page without selecting individual blocks.

{
  "page_id": "page-uuid-here"
}

Delete everything (including child databases/pages):

{
  "page_id": "page-uuid-here",
  "preserve_types": []
}

Returns:

{
  "deleted_count": 15,
  "failed_count": 0
}

create-comment

Add a comment to a page.

{
  "page_id": "page-uuid-here",
  "rich_text": [{ "type": "text", "text": { "content": "This is a comment" } }]
}

create-comment-simple ⭐

Add a comment using Markdown. Simpler than create-comment.

Parameters:

• page_id (required): The ID of the page

• content (required): Comment in Markdown

• discussion_id (optional): Reply to existing thread

{
  "page_id": "page-uuid-here",
  "content": "This is **important** with a [link](https://example.com)"
}

Development

# Install dependencies
pnpm install

# Run in development mode
pnpm dev

# Build for production
pnpm build

# Type check
pnpm typecheck

# Lint
pnpm lint

# Format code
pnpm format

# Run tests
pnpm test

# Run tests in watch mode
pnpm test:watch

License

MIT