Notion SDK for JavaScript

A simple and easy to use client for the Notion API.

Installation

npm install @notionhq/client

Usage

[!NOTE] Use Notion's Getting Started Guide to get set up to use Notion's API.

Import and initialize a client using an integration token or an OAuth access token.

const { Client } = require("@notionhq/client")

// Initializing a client
const notion = new Client({
  auth: process.env.NOTION_TOKEN,
})

Make a request to any Notion API endpoint.

;(async () => {
  const listUsersResponse = await notion.users.list({})
})()

[!NOTE] See the complete list of endpoints in the API reference.

Each method returns a Promise that resolves the response.

console.log(listUsersResponse)

{
  results: [
    {
      object: "user",
      id: "d40e767c-d7af-4b18-a86d-55c61f1e39a4",
      type: "person",
      person: {
        email: "avo@example.org",
      },
      name: "Avocado Lovelace",
      avatar_url:
        "https://secure.notion-static.com/e6a352a8-8381-44d0-a1dc-9ed80e62b53d.jpg",
    },
    // ...
  ]
}

Endpoint parameters are grouped into a single object. You don't need to remember which parameters go in the path, query, or body.

const myPage = await notion.dataSources.query({
  data_source_id: "897e5a76-ae52-4b48-9fdf-e71f5945d1af",
  filter: {
    property: "Landmark",
    rich_text: {
      contains: "Bridge",
    },
  },
})

Handling errors

If the API returns an unsuccessful response, the returned Promise rejects with a APIResponseError.

The error contains properties from the response, and the most helpful is code. You can compare code to the values in the APIErrorCode object to avoid misspelling error codes.

const { Client, APIErrorCode } = require("@notionhq/client")

try {
  const notion = new Client({ auth: process.env.NOTION_TOKEN })
  const myPage = await notion.dataSources.query({
    data_source_id: dataSourceId,
    filter: {
      property: "Landmark",
      rich_text: {
        contains: "Bridge",
      },
    },
  })
} catch (error) {
  if (error.code === APIErrorCode.ObjectNotFound) {
    //
    // For example: handle by asking the user to select a different data source
    //
  } else {
    // Other error handling code
    console.error(error)
  }
}

Logging

The client emits useful information to a logger. By default, it only emits warnings and errors.

If you're debugging an application, and would like the client to log response bodies, set the logLevel option to LogLevel.DEBUG.

const { Client, LogLevel } = require("@notionhq/client")

const notion = new Client({
  auth: process.env.NOTION_TOKEN,
  logLevel: LogLevel.DEBUG,
})

You may also set a custom logger to emit logs to a destination other than stdout. A custom logger is a function which is called with 3 parameters: logLevel, message, and extraInfo. The custom logger should not return a value.

Client options

The Client supports the following options on initialization. These options are all keys in the single constructor parameter.

Option	Default value	Type	Description
auth	undefined	string	Bearer token for authentication. If left undefined, the auth parameter should be set on each request.
logLevel	LogLevel.WARN	LogLevel	Verbosity of logs the instance will produce. By default, logs are written to stdout.
timeoutMs	DEFAULT_TIMEOUT_MS	number	Number of milliseconds to wait before emitting a RequestTimeoutError
baseUrl	DEFAULT_BASE_URL	string	The root URL for sending API requests. This can be changed to test with a mock server.
logger	Log to console	Logger	A custom logging function. This function is only called when the client emits a log that is equal or greater severity than logLevel.
agent	Default node agent	http.Agent	Used to control creation of TCP sockets. A common use is to proxy requests with https-proxy-agent
retry	See constants	RetryOptions	Configuration for automatic retries on rate limits (429) and server errors (500, 503). See Automatic retries below.

Automatic retries

The client automatically retries requests that fail due to rate limiting or transient server errors. By default, it will retry up to 2 times using exponential back-off with jitter.

Retryable errors:

• rate_limited (HTTP 429) - Too many requests; retried for all HTTP methods
• internal_server_error (HTTP 500) - Server error; retried only for GET and DELETE
• service_unavailable (HTTP 503) - Service temporarily unavailable; retried only for GET and DELETE

Server errors (500, 503) are only retried for idempotent HTTP methods (GET, DELETE) to avoid duplicate side effects. Rate limits (429) are retried for all methods since the server explicitly asks clients to retry.

Retry behavior:

• Uses exponential back-off: delays increase with each retry attempt
• Respects the Retry-After header when present (both delta-seconds and HTTP-date formats)
• Adds random jitter to prevent thundering herd problems

Configuration:

const notion = new Client({
  auth: process.env.NOTION_TOKEN,
  retry: {
    maxRetries: 5, // Maximum retry attempts (default: 2)
    initialRetryDelayMs: 500, // Initial delay between retries (default: 1000ms)
    maxRetryDelayMs: 60000, // Maximum delay between retries (default: 60000ms)
  },
})

To disable automatic retries:

const notion = new Client({
  auth: process.env.NOTION_TOKEN,
  retry: false,
})

Constants

The SDK exports named constants for all default values used by the client, as well as useful Notion-specific values. You can import them directly:

const {
  DEFAULT_BASE_URL, // "https://api.notion.com"
  DEFAULT_TIMEOUT_MS, // 60_000
  DEFAULT_MAX_RETRIES, // 2
  DEFAULT_INITIAL_RETRY_DELAY_MS, // 1_000
  DEFAULT_MAX_RETRY_DELAY_MS, // 60_000
  MIN_VIEW_COLUMN_WIDTH, // 32
} = require("@notionhq/client")

MIN_VIEW_COLUMN_WIDTH is the minimum width (in pixels) that a table column can have in the Notion UI. Set a property's width to this value when creating or updating a view to make a column appear collapsed -- useful for checkbox or status-as-checkbox columns:

await notion.views.create({
  database_id: databaseId,
  name: "My view",
  type: "table",
  configuration: {
    table: {
      properties: [
        {
          property_id: checkboxPropId,
          visible: true,
          width: MIN_VIEW_COLUMN_WIDTH,
        },
      ],
    },
  },
})

TypeScript

This package contains type definitions for all request parameters and responses, as well as some useful sub-objects from those entities.

Because errors in TypeScript start with type any or unknown, you should use the isNotionClientError type guard to handle them in a type-safe way. Each NotionClientError type is uniquely identified by its error.code. Codes in the APIErrorCode enum are returned from the server. Codes in the ClientErrorCode enum are produced on the client.

try {
  const response = await notion.dataSources.query({
    /* ... */
  })
} catch (error: unknown) {
  if (isNotionClientError(error)) {
    // error is now strongly typed to NotionClientError
    switch (error.code) {
      case ClientErrorCode.RequestTimeout:
        // ...
        break
      case APIErrorCode.ObjectNotFound:
        // ...
        break
      case APIErrorCode.Unauthorized:
        // ...
        break
      // ...
      default:
        // you could even take advantage of exhaustiveness checking
        assertNever(error.code)
    }
  }
}

Type guards

There are several type guards provided to distinguish between full and partial API responses.

Type guard function	Purpose
isFullPage	Determine whether an object is a full PageObjectResponse
isFullBlock	Determine whether an object is a full BlockObjectResponse
isFullDataSource	Determine whether an object is a full DataSourceObjectResponse
isFullPageOrDataSource	Determine whether an object is a full PageObjectResponse or DataSourceObjectResponse
isFullUser	Determine whether an object is a full UserObjectResponse
isFullComment	Determine whether an object is a full CommentObjectResponse

Here is an example of using a type guard:

const fullOrPartialPages = await notion.dataSources.query({
  data_source_id: "897e5a76-ae52-4b48-9fdf-e71f5945d1af",
})
for (const page of fullOrPartialPages.results) {
  if (!isFullPageOrDataSource(page)) {
    continue
  }
  // The page variable has been narrowed from
  //      PageObjectResponse | PartialPageObjectResponse | DataSourceObjectResponse | PartialDataSourceObjectResponse
  // to
  //      PageObjectResponse | DataSourceObjectResponse.
  console.log("Created at:", page.created_time)
}

Utility functions

This package also exports a few utility functions that are helpful for dealing with any of our paginated APIs.

iteratePaginatedAPI(listFn, firstPageArgs)

This utility turns any paginated API into an async iterator.

Parameters:

• listFn: Any function on the Notion client that represents a paginated API (i.e. accepts start_cursor.) Example: notion.blocks.children.list.
• firstPageArgs: Arguments that should be passed to the API on the first and subsequent calls to the API, for example a block_id.

Returns:

An async iterator over results from the API.

Example:

for await (const block of iteratePaginatedAPI(notion.blocks.children.list, {
  block_id: parentBlockId,
})) {
  // Do something with block.
}

collectPaginatedAPI(listFn, firstPageArgs)

This utility accepts the same arguments as iteratePaginatedAPI, but collects the results into an in-memory array.

Before using this utility, check that the data you are dealing with is small enough to fit in memory.

Parameters:

• listFn: Any function on the Notion client that represents a paginated API (i.e. accepts start_cursor.) Example: notion.blocks.children.list.
• firstPageArgs: Arguments that should be passed to the API on the first and subsequent calls to the API, for example a block_id.

Returns:

An array with results from the API.

Example:

const blocks = await collectPaginatedAPI(notion.blocks.children.list, {
  block_id: parentBlockId,
})
// Do something with blocks.

Custom requests

To make requests directly to a Notion API endpoint instead of using the pre-built families of methods, call the request() method. For example:

// POST /v1/comments
const response = await notion.request({
  path: "comments",
  method: "post",
  body: {
    parent: { page_id: "5c6a28216bb14a7eb6e1c50111515c3d" },
    rich_text: [{ text: { content: "Hello, world!" } }],
  },
  // No `query` params in this example, only `body`.
})

console.log(JSON.stringify(response, null, 2))

The notion.request<ResponseBody>({...}) method is generic; ResponseBody represents the expected type of response object Notion returns for the endpoint you're calling (we don't validate this at runtime; you can pass anything!)

[!TIP] Usually, making custom requests with notion.request() isn't necessary, but can be helpful in some cases, e.g. when upgrading your Notion API version incrementally before upgrading your SDK version. For example, if there's a new or renamed endpoint in the new API version that isn't yet available to call via a dedicated method on Client.

In the above example, the simpler approach is to use await notion.comments.create().

Another customization you can make is to pass your own fetch function to the Client co