bda-research/node-crawler

Crawler v2 : Advanced and Typescript version of node-crawler

Features:

• Server-side DOM & automatic jQuery insertion with Cheerio (default),
• Configurable pool size and retries,
• Control rate limit,
• Priority queue of requests,
• let crawler deal for you with charset detection and conversion,

If you have prior experience with Crawler v1, for fast migration, please proceed to the section Differences and Breaking Changes.

Use with AI agents

This package ships an agent skill that teaches AI agents (e.g. OpenClaw, Hermes, Claude Code, Codex etc) how to crawl with crawler — queue vs. direct requests, rate limiting, proxy rotation, Cheerio parsing, and common pitfalls.

Recommended — install from ClawHub: clawhub.ai/mike442144/node-crawler

openclaw skills install node-crawler

Fallback — download the skill bundle from the latest release and unzip it into your agent's skills directory:

curl -L -o node-crawler-skill.zip https://github.com/bda-research/node-crawler/releases/latest/download/node-crawler-skill.zip
unzip node-crawler-skill.zip -d .openclaw/skills/

Let your agent install it — paste this prompt:

Install the node-crawler skill so you can use it for large-scale web crawling: download https://github.com/bda-research/node-crawler/releases/latest/download/node-crawler-skill.zip, unzip it, and place the node-crawler/ folder in your skills directory (e.g. .openclaw/skills/, .claude/skills/, or wherever your runtime loads skills from).

The agent loads it automatically when a task involves large-scale scraping or crawling.

Quick start

Install

Requires Node.js 22 or above.

$ npm install crawler

Warning: Given the dependencies involved (Especially migrating from request to got) , Crawler v2 has been designed as a native ESM and no longer offers a CommonJS export. We would also like to recommend that you convert to ESM. Note that making this transition is generally not too difficult.If you have a large codebase built with Crawler v1, you can upgrade to v2.0.3-beta (using npm install crawler@beta), which supports both ESM and CommonJS builds.Please note that code previously using the "body" parameter to send form data in POST requests will need to be updated to use "form" even in the beta version.

Usage

Execute asynchronously via custom options

import Crawler from "crawler";

const c = new Crawler({
    maxConnections: 10,
    // This will be called for each crawled page
    callback: (error, res, done) => {
        if (error) {
            console.log(error);
        } else {
            const $ = res.$;
            // $ is Cheerio by default
            //a lean implementation of core jQuery designed specifically for the server
            console.log($("title").text());
        }
        done();
    },
});

// Add just one URL to queue, with default callback
c.add("http://www.amazon.com");

// Add a list of URLs
c.add(["http://www.google.com/", "http://www.yahoo.com"]);

// Add URLs with custom callbacks & parameters
c.add([
    {
        url: "http://parishackers.org/",
        jQuery: false,

        // The global callback won't be called
        callback: (error, res, done) => {
            if (error) {
                console.log(error);
            } else {
                console.log("Grabbed", res.body.length, "bytes");
            }
            done();
        },
    },
]);

// Add some HTML code directly without grabbing (mostly for tests)
c.add([
    {
        html: "<title>This is a test</title>",
    },
]);

please refer to options for detail.

Slow down

Use rateLimit to slow down when you are visiting web sites.

import Crawler from "crawler";

const c = new Crawler({
    rateLimit: 1000, // `maxConnections` will be forced to 1
    callback: (err, res, done) => {
        console.log(res.$("title").text());
        done();
    },
});

c.add(tasks); //between two tasks, minimum time gap is 1000 (ms)

Custom parameters

Sometimes you have to access variables from previous request/response session, what should you do is passing parameters in options.userParams :

c.add({
    url: "http://www.google.com",
    userParams: {
        parameter1: "value1",
        parameter2: "value2",
        parameter3: "value3",
    },
});

then access them in callback via res.options

console.log(res.options.userParams);

Raw body

If you are downloading files like image, pdf, word etc, you have to save the raw response body which means Crawler shouldn't convert it to string. To make it happen, you need to set encoding to null

import Crawler from "crawler";
import fs from "fs";

const c = new Crawler({
    encoding: null,
    jQuery: false, // set false to suppress warning message.
    callback: (err, res, done) => {
        if (err) {
            console.error(err.stack);
        } else {
            fs.createWriteStream(res.options.userParams.filename).write(res.body);
        }
        done();
    },
});

c.add({
    url: "https://raw.githubusercontent.com/bda-research/node-crawler/master/crawler_primary.png",
    userParams: {
        filename: "crawler.png",
    },
});

preRequest

If you want to do something either synchronously or asynchronously before each request, you can try the code below. Note that direct requests won't trigger preRequest.

import Crawler from "crawler";

const c = new Crawler({
    preRequest: (options, done) => {
        // 'options' here is not the 'options' you pass to 'c.queue', instead, it's the options that is going to be passed to 'request' module
        console.log(options);
        // when done is called, the request will start
        done();
    },
    callback: (err, res, done) => {
        if (err) {
            console.log(err);
        } else {
            console.log(res.statusCode);
        }
    },
});

c.add({
    url: "http://www.google.com",
    // this will override the 'preRequest' defined in crawler
    preRequest: (options, done) => {
        setTimeout(() => {
            console.log(options);
            done();
        }, 1000);
    },
});

Direct request

Support both Promise and callback

import Crawler from "crawler";

const crawler = new Crawler();

// When using directly "send", the preRequest won't be called and the "Event:request" won't be triggered
const response = await crawler.send("https://github.com/");
console.log(response.options);
// console.log(response.body);

crawler.send({
    url: "https://github.com/",
    // When calling `send`, `callback` must be defined explicitly, with two arguments `error` and `response`
    callback: (error, response) => {
        if (error) {
            console.error(error);
        } else {
            console.log("Hello World!");
        }
    },
});

Table

• Content
  • Work with Http2
  • Work with rateLimiters
  • Class: Crawler
    • Event: 'schedule'
    • Event: 'limiterChange'
    • Event: 'request'
    • Event: 'drain'
    • crawler.add(url|options)
    • crawler.queueSize
  • Options
    • Global only options
      • silence
      • maxConnections
      • priorityLevels
      • rateLimit
      • skipDuplicates
      • homogeneous
      • userAgents
    • Crawler General options
      • url | method | headers | body | searchParams...
      • forceUTF8
      • jQuery
      • encoding
      • rateLimiterId
      • retries
      • retryInterval
      • timeout
      • priority
      • skipEventRequest
      • html
      • proxies
      • proxy
      • http2
      • autoSelectFamily
      • autoSelectFamilyAttemptTimeout
      • referer
      • userParams
      • preRequest
      • Callback
  • Work with Cheerio
• Differences and Breaking Changes
  • renaming
    • Crawler Options
    • Origin Request Options
  • Behavior Changes
• How to test

Content

Work with Http2

Now we offer hassle-free support for using HTTP/2: just set http2 to true, and Crawler will operate as smoothly as with HTTP (including proxies).

Note: As most developers using this library with proxies also work with Charles, it is expected to set rejectAuthority to false in order to prevent the so-called 'self-signed certificate' errors."

crawler.send({
    url: "https://nghttp2.org/httpbin/status/200",
    method: "GET",
    http2: true,
    callback: (error, response) => {
        if (error) {
            console.error(error);
        }
        console.log(`inside callback`);
        console.log(response.body);
    },
});

Work with rateLimiters

Control the rate limit. All tasks submit to a rateLimiter will abide the rateLimit and maxConnections restrictions of the limiter. rateLimit is the minimum time gap between two tasks. maxConnections is the maximum number of tasks that can be running at the same time. rateLimiters are independent of each other. One common use case is setting different rateLimiters for different proxies. One thing is worth noticing, when rateLimit is set to a non-zero value, maxConnections will be forced to 1.

import Crawler from "crawler";

const c = new Crawler({
    rateLimit: 2000,
    maxConnections: 1,
    callback: (error, res, done) => {
        if (error) {
            console.log(error);
        } else {
            const $ = res.$;
            console.log($("title").text());
        }
        done();
    },
});

// if you want to crawl some website with 2000ms gap between requests
c.add("http://www.somewebsite.com/page/1");
c.add("http://www.somewebsite.com/page/2");
c.add("http://www.somewebsite.com/page/3");

// if you want to crawl some website using proxy with 2000ms gap between requests for each proxy
c.add({
    url: "http://www.somewebsite.com/page/1",
    rateLimiterId: 1,
    proxy: "proxy_1",
});
c.add({
    url: "http://www.somewebsite.com/page/2",
    rateLimiterId: 2,
    proxy: "proxy_2",
});
c.add({
    url: "http://www.somewebsite.com/page/3",
    rateLimiterId: 3,
    proxy: "proxy_3",
});
c.add({
    url: "http://www.somewebsite.com/page/4",
    rateLimiterId: 4,
    proxy: "proxy_1",
});

Normally, all ratelimiters instances in the limiter cluster of crawler are instantiated with options specified in crawler constructor. You can change property of any rateLimiter by calling the code below. Currently, we only support changing property 'rateLimit' of it. Note that the default rateLimiter can be accessed by crawler.setLimiter(0, "rateLimit", 1000);. We strongly recommend that you leave limiters unchanged after their instantiation unless you know clearly what you are doing.

const crawler = new Crawler();
crawler.setLimiter(0, "rateLimit", 1000);

Class: Crawler

Event: 'schedule'

• options

Emitted when a task is being added to scheduler.

crawler.on("schedule", options => {
    options.proxy = "http://proxy:port";
});

Event: 'limiterChange'

• options
• rateLimiterId : number

Emitted when limiter has been changed.

Event: 'request'

• options

Emitted when crawler is ready to send a request.

If you are going to modify options at last stage before requesting, just listen on it.

crawler.on("request", options => {
    options.searchParams.timestamp = new Date().getTime();
});

Event: 'drain'

Emitted when queue is empty.

crawler.on("drain", () => {
    // For example, release a connection to database.
    db.end(); // close connection to MySQL
});

crawler.add(url|options)

• url | options

Add a task to queue and wait for it to be executed.

crawler.queueSize

• Number

Size of queue, read-only

Options

You can pass these options to the Crawler() constructor if you want them to be global or as items in the crawler.add() calls if you want them to be specific to that item (overwriting global options)

• For using easily, simply passing a url string as Options is also accepted.
• Options can also be an array composed of multiple options, in which case multiple tasks will be added at once.
• When constructing options, all native got options are accepted and passed through directly. Additionally, the options are tailored to process only those parameters that are identifiable by the Crawler.

Global only options

silence

• Type: boolean
• Default : false
• If true, the crawler will mute all warning and error messages. The request error will be still reported.

maxConnections

• Type: number
• Default : 10
• The maximum number of requests that can be sent simultaneously. If the value is 10, the crawler will send at most 10 requests at the same time.

priorityLevels

• Type: number
• Default : 10
• The number of levels of priority. Can be only assigned at the beginning.

rateLimit

• Type: number

• Default : 0

• 1000 means 1000 milliseconds delay between after the first request.

• Note: This options is list as global only options because it will be set as the "default rateLimit value". This value is bound to a specific rate limiter and can only be modified through the crawler.setLimiter method. Please avoid passing redundant rateLimit property in local requests; instead, use options.rateLimiterId to specify a particular limiter.

• Example:

  crawler.on("schedule", options => {
      options.rateLimiterId = Math.floor(Math.random() * 15);
  });
  

skipDuplicates

• Type: boolean
• Default : false
• If true, the crawler will skip duplicate tasks. If the task is already in the queue, the crawler will not add it again.

homogeneous

• Type: boolean
• Default : false
• If true, the crawler will dynamically reallocate the tasks within the queue blocked due to header blocking to other queues.

userAgents

• Type: string | string[]
• Default : undefined
• If passed, the crawler will rotate the user agent for each request. The "userAgents" option must be an array if activated.

Crawler General options

url | method | headers | body | searchParams...

• Same as the options of options

forceUTF8

• Type: boolean
• Default : false
• If true, the crawler will detect the charset from the HTTP headers or the meta tag in the HTML and convert it to UTF-8 if necessary.

jQuery

• Type: boolean
• Default : true
• If true, the crawler will use the cheerio library to parse the HTML content.

encoding

• Type: string
• Default : 'utf8'
• The encoding of the response body.

rateLimiterId

• Type: number
• Default : 0
• The rateLimiter ID.

retries

• Type: number
• Default : 2
• The number of retries if the request fails.

retryInterval

• Type: number
• Default : 3000
• The number of milliseconds to wait before retrying.

timeout

• Type: number
• Default : 20000
• The number of milliseconds to wait before the request times out.

priority

• Type: number
• Default : 5
• The priority of the request.

skipEventRequest

• Type: boolean
• Default : false
• If true, the crawler will not trigger the 'request' event.

html

• Type: boolean
• Default : true
• If true, the crawler will parse the response body as HTML.

isJson

• Type: boolean
• Default : false
• If true, the crawler will parse the response body as JSON. This will also force the encoding to UTF-8 and disable jQuery (cheerio) parsing.

proxies

• Type: string[]
• Default : []
• The list of proxies. If passed, the proxy will be rotated by requests.
• Warning: It is recommended to avoid the usage of "proxies", better to use the following method instead. (Probably you can understand why...)

const ProxyManager = {
    index: 0,
    proxies: JSON.parse(fs.readFileSync("../proxies.json")),
    setProxy: function (options) {
        let proxy = this.proxies[this.index];
        this.index = ++this.index % this.proxies.length;
        options.proxy = proxy;
        options.rateLimiterId = Math.floor(Math.random() * 15);
    },
};

crawler.on("schedule", options => {
    // options.proxy = "http://127.0.0.1:8000";
    ProxyManager.setProxy(options);
});

proxy

• Type: string
• Default : undefined
• The proxy to use. The priority is higher than the "proxies" option.

http2

• Type: boolean
• Default : false
• If true, the request will be sent in the HTTP/2 protocol.

autoSelectFamily

• Type: boolean
• Default : true (Node default)
• Controls Node's "Happy Eyeballs", which races IPv6 and IPv4 when a host has both. Set to false on networks with broken or slow IPv6 to use a single family and avoid spurious ETIMEDOUT errors.
• Note: Node applies this as a process-wide default, so it affects all in-flight requests.

autoSelectFamilyAttemptTimeout

• Type: number
• Default : 250 (Node default)
• How long, in milliseconds, "Happy Eyeballs" waits for one address before trying the next. Raise it (e.g. 5000) for high-latency hosts whose connection can't finish within 250ms, which otherwise fails with ETIMEDOUT before the working address responds.
• Note: Node applies this as a process-wide default, so it affects all in-flight requests.

referer

• Type: string
• Default : undefined
• If truthy, sets the HTTP referer header.

userParams

• Type: unknown
• Default : undefined
• The user parameters. You can access them in the callback via res.options.

preRequest

• Type: (options, done) => unknown
• Default : undefined
• The function to be called before each request. Only works for the crawler.add method.

Callback

• Type: (error, response, done) => unknown

• Function that will be called after a request was completed

  • error: Error catched by the crawler
  • response : A response of standard IncomingMessage includes $ and options
    • response.options: Options of this task
    • response.$: jQuery Selector A selector for html or xml document.
    • response.statusCode: Number HTTP status code. E.G.200
    • response.body: Buffer | String | JSON HTTP response content which could be a html page, plain text or xml document e.g.
    • response.headers: HTTP response headers
  • done : The function must be called when you've done your work in callback. This is the only way to tell the crawler that the task is finished.

Work with Cheerio

Crawler by default use Cheerio. We are temporarily no longer supporting jsdom for certain reasons, may be later.

Differences and Breaking Changes

renaming

Options list here are renamed but most of the old ones are still supported for backward compatibility.

Crawler Options

options.priorityRange → options.priorityLevels

options.uri → options.url

options.json → options.isJson (Boolean. The "json" option is now work completely different.)

options.limiter → options.rateLimiterId

options.retryTimeout → options.retryInterval

crawler.direct → crawler.send

crawler.queue → crawler.add

crawler.setLimiterProperty → crawler.setLimiter

Origin Request Options

incomingEncoding → encoding

qs → searchParams

strictSSL → rejectUnauthorized

gzip → decompress

jar → cookieJar (accepts tough-cookie jar)

jsonReviver → parseJson

jsonReplacer → stringifyJson

Behavior Changes

• default retries: 3 => 2

Some practices that were acceptable and offen used in version 1 but not in version 2：

• use "body" as the POST form => Please use "form" instead. For more, see options .
• add custom options on request options => Not allowed. Only options.userParams could pass through the response.
• We are temporarily no longer supporting jsdom for certain reasons.

How to test

Crawler uses nock to mock http request, thus testing no longer relying on http server.

$ pnpm test