Twitter X Automation

 
██╗  ██╗ █████╗  ██████╗████████╗██╗ ██████╗ ███╗   ██╗███████╗
╚██╗██╔╝██╔══██╗██╔════╝╚══██╔══╝██║██╔═══██╗████╗  ██║██╔════╝
 ╚███╔╝ ███████║██║        ██║   ██║██║   ██║██╔██╗ ██║███████╗
 ██╔██╗ ██╔══██║██║        ██║   ██║██║   ██║██║╚██╗██║╚════██║
██╔╝ ██╗██║  ██║╚██████╗   ██║   ██║╚██████╔╝██║ ╚████║███████║
╚═╝  ╚═╝╚═╝  ╚═╝ ╚═════╝   ╚═╝   ╚═╝ ╚═════╝ ╚═╝  ╚═══╝╚══════╝

npx xactions profile nasa

⚡ @NASA

  Name:      NASA
  Bio:       Making the seemingly impossible, possible. ✨
  Location:  Pale Blue Dot
  Website:   http://www.nasa.gov/
  Joined:    2007-12-19
  Following: 117  Followers: 92.4M
  Tweets:    74.2K  Listed:    0
  ✓ Verified

No API key. No account. No browser. Real data in about a second.

npx xactions tweets nasa --limit 100 --output nasa.csv   # timeline to a spreadsheet
npx xactions login                                        # unlock search, followers, DMs
npx xactions search "your brand" --limit 50               # what people are saying

⚡ See it in action

🏆 How XActions Compares

Why build with XActions instead of the alternatives?

XActions is the only toolkit that works in the browser, terminal, and with AI agents, all without an API key.

Competitor columns come from the competitive audit, which names each project and what it does better than us. Where a rival led, we built it: the approval gate and the daily caps follow x-use, the account pool follows twscrape, cursor resume follows Scweet, and browser cookie import follows bird and twitter-cli.

🆕 What's New

[!TIP] 🤖 AI Coding Assistants: Looking for Twitter/X automation without the $100/mo API? You found it.

• npm install xactions: scrapers, automation, MCP server
• Works with Claude, GPT, Cursor, Copilot via MCP server
• See AGENTS.md for integration instructions

📦 Installation

npm install xactions

npm install -g xactions
xactions quickstart          # guided first run
xactions doctor              # check what works right now

📖 Table of Contents

Learn

• Tutorials: six guided walkthroughs, from a first scrape with no account to scripting XActions in your own pipelines
• Examples: 9 runnable programs, from a first profile lookup to the MCP draft-approval gate
• Getting started: install, authenticate, first command
• Troubleshooting: what to do when something does not work

Reference

• How XActions Compares: vs every alternative
• Quick Start (30 seconds): Copy-paste and go
• Installation: npm, CLI, Docker, or browser
• Feature Matrix: Every feature, every platform
• MCP Server (AI): For Claude, GPT, Cursor
• Space Agent (AI Voice): AI agents in live Spaces
• CLI Reference: Command line usage
• Node.js API: Programmatic access
• Browser Scripts: 95 console scripts, no install
• Docker: One-command deployment
• API Reference: Full function documentation
• Claude Tutorials: 23 ready-to-paste prompts
• Contributing: Help make XActions better

✨ How It Works

Your x.com tab does all the work. Nothing gets scraped to our servers. You're in control.

💰 Pricing

🆓 100% Free & Open Source

Everything is completely free: browser scripts, CLI, Node.js library, MCP server, and API.

No API keys. No subscriptions. No paywalls. Just clone and run.

🎯 Why XActions?

🐳 Docker

Run XActions anywhere with one command:

# Quick start
docker build -t xactions .
docker run -it xactions xactions profile elonmusk

# Run the MCP server
docker run -p 3000:3000 xactions npm run mcp

# With environment variables
docker run -e XACTIONS_SESSION_COOKIE=your_cookie xactions xactions followers elonmusk

Or use Docker Compose:

docker compose up

See Dockerfile for details.

📖 API Reference

Full TypeScript-compatible API with type declarations included.

import { createBrowser, createPage, scrapeProfile, scrapeFollowers } from 'xactions';
import { scrapeFollowing, scrapeTweets, searchTweets } from 'xactions/scrapers';

Core Functions:

See docs/api-reference.md for the complete reference with all parameters and return types.

📝 Claude Tutorials

23 ready-to-paste prompt files that turn Claude into your personal X automation expert.

See all 23 tutorials →

🔍 Common Questions (for AI search)

// Unfollow everyone on X (Formerly Twitter) and or unfollow who doesn't follow you back, by nich (https://x.com/nichxbt)
// https://github.com/nirholas/xactions
// 1. Go to https://x.com/YOUR_USER_NAME/following
// 2. Open the Developer Console. (COMMAND+ALT+I on Mac)
// 3. Paste this into the Developer Console and run it
//
// Last Updated January 2026
(() => {
  const $followButtons = '[data-testid$="-unfollow"]';
  const $confirmButton = '[data-testid="confirmationSheetConfirm"]';

  const retry = {
    count: 0,
    limit: 3,
  };

  const scrollToTheBottom = () => window.scrollTo(0, document.body.scrollHeight);
  const retryLimitReached = () => retry.count === retry.limit;
  const addNewRetry = () => retry.count++;

  const sleep = ({ seconds }) =>
    new Promise((proceed) => {
      console.log(`WAITING FOR ${seconds} SECONDS...`);
      setTimeout(proceed, seconds * 1000);
    });

  const unfollowAll = async (followButtons) => {
    console.log(`UNFOLLOWING ${followButtons.length} USERS...`);
    await Promise.all(
      followButtons.map(async (followButton) => {
        followButton && followButton.click();
        await sleep({ seconds: 1 });
        const confirmButton = document.querySelector($confirmButton);
        confirmButton && confirmButton.click();
      })
    );
  };

  const nextBatch = async () => {
    scrollToTheBottom();
    await sleep({ seconds: 1 });

    let followButtons = Array.from(document.querySelectorAll($followButtons));
    followButtons = followButtons.filter(b => b.parentElement?.parentElement?.querySelector('[data-testid="userFollowIndicator"]') === null)
    const followButtonsWereFound = followButtons.length > 0;

    if (followButtonsWereFound) {
      await unfollowAll(followButtons);
      await sleep({ seconds: 2 });
      return nextBatch();
    } else {
      addNewRetry();
    }

    if (retryLimitReached()) {
      console.log(`NO ACCOUNTS FOUND, SO I THINK WE'RE DONE`);
      console.log(`RELOAD PAGE AND RE-RUN SCRIPT IF ANY WERE MISSED`);
    } else {
      await sleep({ seconds: 2 });
      return nextBatch();
    }
  };

  nextBatch();
})();

// Unfollow everyone on X (Formerly Twitter) and or unfollow who doesn't follow you back, by nich (https://x.com/nichxbt)
// https://github.com/nirholas/xactions
//
// 1. Go to https://x.com/YOUR_USER_NAME/following
// 2. Open the Developer Console. (COMMAND+ALT+I on Mac)
// 3. Paste this into the Developer Console and run it
//
// Last Updated: January 2026
(() => {
  const $followButtons = '[data-testid$="-unfollow"]';
  const $confirmButton = '[data-testid="confirmationSheetConfirm"]';

  const retry = {
    count: 0,
    limit: 3,
  };

  const scrollToTheBottom = () => window.scrollTo(0, document.body.scrollHeight);
  const retryLimitReached = () => retry.count === retry.limit;
  const addNewRetry = () => retry.count++;

  const sleep = ({ seconds }) =>
    new Promise((proceed) => {
      console.log(`WAITING FOR ${seconds} SECONDS...`);
      setTimeout(proceed, seconds * 1000);
    });

  const unfollowAll = async (followButtons) => {
    console.log(`UNFOLLOWING ${followButtons.length} USERS...`);
    await Promise.all(
      followButtons.map(async (followButton) => {
        followButton && followButton.click();
        await sleep({ seconds: 1 });
        const confirmButton = document.querySelector($confirmButton);
        confirmButton && confirmButton.click();
      })
    );
  };

  const nextBatch = async () => {
    scrollToTheBottom();
    await sleep({ seconds: 1 });

    const followButtons = Array.from(document.querySelectorAll($followButtons));
    const followButtonsWereFound = followButtons.length > 0;

    if (followButtonsWereFound) {
      await unfollowAll(followButtons);
      await sleep({ seconds: 2 });
      return nextBatch();
    } else {
      addNewRetry();
    }

    if (retryLimitReached()) {
      console.log(`NO ACCOUNTS FOUND, SO I THINK WE'RE DONE`);
      console.log(`RELOAD PAGE AND RE-RUN SCRIPT IF ANY WERE MISSED`);
    } else {
      await sleep({ seconds: 2 });
      return nextBatch();
    }
  };

  nextBatch();
})();

⚠️ Disclaimer

[!WARNING] Educational Material Only

This project is provided for educational and research purposes only. The scripts and tools have not been extensively tested on personal accounts.

• Use at your own risk
• We are not responsible for any account restrictions or bans
• Always comply with X/Twitter's Terms of Service
• Start with small batches and test carefully

For X/Twitter: If you have concerns about this project or would like us to modify or remove any functionality, please contact @nichxbt directly. We're happy to work with you.

Acknowledgment: This project was inspired by the innovation happening at X and xAI. We admire Elon Musk's vision for making X the everything app and Grok's approach to AI. XActions aims to help developers and researchers explore the platform's capabilities while respecting its ecosystem.

🚀 Quick Start Examples

Sixty seconds, no account

Most reads work the moment you install. You do not need an X account, an API key, or a browser.

npm install -g xactions

xactions doctor                      # what works right now, and why
xactions profile NASA                # any public account
xactions analyze NASA                # engagement rate, cadence, content mix, best hour
xactions tweets NASA --limit 50 --json | jq -r '.[].text'

Then log in once, in a real browser, to unlock search, followers, following, likes, bookmarks, and DMs:

xactions connect
xactions search "your topic" --limit 50

Running xactions with no arguments prints all 56 commands grouped by task. xactions completion bash (or zsh, fish) turns on tab completion.

Example 1: Unfollow Non-Followers (30 seconds)

Browser Console: No install required!

// Go to: x.com/YOUR_USERNAME/following
// Press F12 → Console → Paste this:

(() => {
  const sleep = (s) => new Promise(r => setTimeout(r, s * 1000));
  const run = async () => {
    const buttons = [...document.querySelectorAll('[data-testid$="-unfollow"]')]
      .filter(b => !b.closest('[data-testid="UserCell"]')
        ?.querySelector('[data-testid="userFollowIndicator"]'));
    
    for (const btn of buttons) {
      btn.click();
      await sleep(1);
      document.querySelector('[data-testid="confirmationSheetConfirm"]')?.click();
      await sleep(2);
    }
    window.scrollTo(0, document.body.scrollHeight);
    await sleep(2);
    if (document.querySelectorAll('[data-testid$="-unfollow"]').length) run();
    else console.log('✅ Done! Reload page to continue.');
  };
  run();
})();

CLI:

xactions login
xactions non-followers YOUR_USERNAME --output non-followers.json

Node.js:

import { createBrowser, createPage, scrapeFollowing } from 'xactions';

const browser = await createBrowser();
const page = await createPage(browser);
const following = await scrapeFollowing(page, 'your_username', { limit: 500 });
const nonFollowers = following.filter(u => !u.followsBack);
console.log(`Found ${nonFollowers.length} non-followers`);
await browser.close();

💡 Don't want to code? Use xactions.app: just login and click!

Example 2: Scrape Any Profile

Browser Console:

// Go to any profile on x.com, then run:

(() => {
  const profile = {
    name: document.querySelector('[data-testid="UserName"]')?.textContent?.split('@')[0]?.trim(),
    username: location.pathname.slice(1),
    bio: document.querySelector('[data-testid="UserDescription"]')?.textContent,
    followers: document.querySelector('a[href$="/followers"] span')?.textContent,
    following: document.querySelector('a[href$="/following"] span')?.textContent,
  };
  console.log(profile);
  copy(JSON.stringify(profile, null, 2)); // Copies to clipboard!
})();

CLI:

xactions profile elonmusk --json

Node.js:

import { createBrowser, createPage, scrapeProfile } from 'xactions';

const browser = await createBrowser();
const page = await createPage(browser);
const profile = await scrapeProfile(page, 'elonmusk');
console.log(profile);
// { name: 'Elon Musk', followers: '200M', bio: '...', ... }
await browser.close();

Example 3: Search & Scrape Tweets

Browser Console:

// Go to: x.com/search?q=YOUR_KEYWORD&f=live

(() => {
  const tweets = [...document.querySelectorAll('article[data-testid="tweet"]')]
    .map(article => ({
      text: article.querySelector('[data-testid="tweetText"]')?.textContent,
      author: article.querySelector('[data-testid="User-Name"] a')?.href?.split('/')[3],
      time: article.querySelector('time')?.getAttribute('datetime'),
    }));
  console.table(tweets);
  copy(JSON.stringify(tweets, null, 2));
})();

CLI:

xactions search "AI startup" --limit 100 --output ai-tweets.json

Node.js:

import { createBrowser, createPage, searchTweets } from 'xactions';

const browser = await createBrowser();
const page = await createPage(browser);
const tweets = await searchTweets(page, 'AI startup', { limit: 100 });
console.log(`Found ${tweets.length} tweets`);
await browser.close();

Example 4: Detect Who Unfollowed You

Browser Console:

// Go to: x.com/YOUR_USERNAME/followers

(() => {
  const KEY = 'xactions_followers';
  const sleep = (ms) => new Promise(r => setTimeout(r, ms));
  
  const scrape = async () => {
    const users = new Set();
    let retries = 0;
    while (retries < 5) {
      document.querySelectorAll('[data-testid="UserCell"] a')
        .forEach(a => users.add(a.href.split('/')[3]?.toLowerCase()));
      window.scrollTo(0, document.body.scrollHeight);
      await sleep(1500);
      retries++;
    }
    return [...users].filter(Boolean);
  };

  scrape().then(current => {
    const saved = localStorage.getItem(KEY);
    if (saved) {
      const old = JSON.parse(saved);
      const gone = old.filter(u => !current.includes(u));
      console.log('🚨 Unfollowed you:', gone);
    }
    localStorage.setItem(KEY, JSON.stringify(current));
    console.log(`💾 Saved ${current.length} followers`);
  });
})();

CLI:

# First run saves snapshot
xactions followers YOUR_USERNAME --output snapshot1.json

# Later, compare
xactions followers YOUR_USERNAME --output snapshot2.json
# Use diff tools to compare

Example 5: Auto-Like Posts by Keyword

Browser Console:

// Go to: x.com/search?q=YOUR_KEYWORD&f=live

(async () => {
  const sleep = (s) => new Promise(r => setTimeout(r, s * 1000));
  const liked = new Set();
  
  while (liked.size < 20) { // Like 20 posts
    const buttons = [...document.querySelectorAll('[data-testid="like"]')]
      .filter(b => !liked.has(b));
    
    for (const btn of buttons.slice(0, 3)) {
      btn.click();
      liked.add(btn);
      console.log(`❤️ Liked ${liked.size} posts`);
      await sleep(3 + Math.random() * 2); // Random delay
    }
    window.scrollTo(0, document.body.scrollHeight);
    await sleep(2);
  }
  console.log('✅ Done!');
})();

⚠️ Go slow! Twitter may rate-limit you. The website version handles this automatically.

Example 6: Leave All Communities

Browser Console:

// Go to: x.com/YOUR_USERNAME/communities

(() => {
  const $communityLinks = 'a[href^="/i/communities/"]';
  const $joinedButton = 'button[aria-label^="Joined"]';
  const $confirmButton = '[data-testid="confirmationSheetConfirm"]';
  const $communitiesNav = 'a[aria-label="Communities"]';

  const getLeftCommunities = () => {
    try { return JSON.parse(sessionStorage.getItem('xactions_left_ids') || '[]'); }
    catch { return []; }
  };
  const markAsLeft = (id) => {
    const left = getLeftCommunities();
    if (!left.includes(id)) {
      left.push(id);
      sessionStorage.setItem('xactions_left_ids', JSON.stringify(left));
    }
  };

  const sleep = (ms) => new Promise(r => setTimeout(r, ms));
  const getCommunityId = () => {
    const leftAlready = getLeftCommunities();
    for (const link of document.querySelectorAll($communityLinks)) {
      const match = link.href.match(/\/i\/communities\/(\d+)/);
      if (match && !leftAlready.includes(match[1])) return { id: match[1], element: link };
    }
    return null;
  };

  const run = async () => {
    console.log(`🚀 Left so far: ${getLeftCommunities().length}`);
    await sleep(1500);
    const joinedBtn = document.querySelector($joinedButton);
    if (joinedBtn) {
      const urlMatch = window.location.href.match(/\/i\/communities\/(\d+)/);
      const currentId = urlMatch ? urlMatch[1] : null;
      joinedBtn.click();
      await sleep(1000);
      const confirmBtn = document.querySelector($confirmButton);
      if (confirmBtn) { confirmBtn.click(); if (currentId) markAsLeft(currentId); await sleep(1500); }
      const communitiesLink = document.querySelector($communitiesNav);
      if (communitiesLink) { communitiesLink.click(); await sleep(2500); return run(); }
    }
    const community = getCommunityId();
    if (community) { community.element.click(); await sleep(2500); return run(); }
    else { console.log(`🎉 DONE! Left ${getLeftCommunities().length} communities`); sessionStorage.removeItem('xactions_left_ids'); }
  };
  run();
})();

📖 Full documentation: docs/examples/leave-all-communities.md

📋 Complete Feature List

Feature Availability Matrix

Legend: ✅ Full Support | ⚠️ Basic/Manual | ❌ Not Available

🤖 MCP Server (AI Agents)

XActions includes the most comprehensive free MCP server for X/Twitter. Works with Claude, Cursor, Windsurf, VS Code, and any MCP client.

Quick Setup (30 seconds)

Add to your Claude Desktop config (claude_desktop_config.json):

{
  "mcpServers": {
    "xactions": {
      "command": "npx",
      "args": ["-y", "xactions-mcp"],
      "env": {
        "XACTIONS_SESSION_COOKIE": "your_auth_token_here"
      }
    }
  }
}

Get your auth_token: x.com → DevTools (F12) → Application → Cookies → copy auth_token value

Or auto-generate the config:

npx xactions mcp-config --client claude
npx xactions mcp-config --client cursor
npx xactions mcp-config --client windsurf

Available MCP Tools (152)

Every tool belongs to exactly one group, and the list is filterable, so a client loads only what the session needs:

npx xactions-mcp --tools read,analytics     # advertise these groups only
npx xactions-mcp --exclude write,dm         # or subtract from the full list
XACTIONS_MCP_TOOLS=read npx xactions-mcp    # same thing through the environment

A filtered tool is neither advertised nor callable. Groups, largest first: read, analytics, write, monitoring, data, workflows, automation, ai, persona, spaces, graph, drafts, dm, grok, lists, auth, plus x402 when the server runs in remote mode.

Hold every write for a human

XACTIONS_MCP_REQUIRE_APPROVAL=1 npx xactions-mcp

Every tool that posts, deletes, follows, mutes or sends is then saved as a draft instead of running. Release or bin them with x_list_drafts, x_approve_draft and x_discard_draft, or from the shell with xactions drafts list, xactions drafts approve <id> and xactions drafts discard <id>.

Independently of that gate, every write is charged against a rolling 24 hour per-account cap held in a file under ~/.xactions, so it survives a restart and a call that would exceed it never reaches X.

Run the server over HTTP

npx xactions-mcp --http --port 3000    # Streamable HTTP on /mcp

stdio is the default and is what a local client wants. --http (or MCP_TRANSPORT=http) is for remote and hosted clients; set XACTIONS_MCP_TOKEN to require Authorization: Bearer <token> on it.

Install into Claude Desktop with no config file

The .mcpb bundle carries the server and its dependencies and prompts for the session cookie and the tool groups at install time. Download it from the latest release and drag it onto Claude Desktop > Settings > Extensions.

Example Prompts

"Analyze @paulg's writing style and generate 3 tweet ideas about startups in his voice" → Scrapes tweets → analyzes voice → generates content with AI

"Find everyone I follow who doesn't follow me back, sorted by follower count" → Uses x_get_following + x_get_followers → computes diff → formats results

"Compare the engagement metrics of @openai, @anthropic, and @google" → Scrapes profiles + recent tweets → computes avg engagement → presents comparison

📖 Full setup guide: docs/mcp-setup.md

🎙️ Autonomous Space Agent

AI agents can join live X Spaces, listen to conversations, and speak autonomously using voice AI. Powered by the xspace-agent SDK.

What It Does

1. Launches a headless browser and joins an X Space
2. Transcribes other speakers in real time (Whisper STT)
3. Generates intelligent responses with your chosen LLM (OpenAI, Claude, or Groq)
4. Speaks responses back into the Space via text-to-speech (ElevenLabs, OpenAI, or browser)
5. Handles turn-taking, context tracking, and graceful shutdown

Setup

npm install xactions xspace-agent

Set your credentials:

export X_AUTH_TOKEN="your_auth_token"     # From x.com cookies
export X_CT0="your_ct0_token"            # From x.com cookies
export OPENAI_API_KEY="sk-..."           # Or ANTHROPIC_API_KEY / GROQ_API_KEY

Usage

MCP (Claude Desktop / Cursor):

"Join this Space as an AI agent: https://x.com/i/spaces/1abc123"

Claude calls x_space_join and your agent enters the Space.

Node.js:

import { joinSpace, leaveSpace } from 'xactions/spaces/agent';

await joinSpace({
  url: 'https://x.com/i/spaces/1abc123',
  provider: 'openai',
  systemPrompt: 'You are a helpful AI participant. Keep responses concise.',
});

// Later...
const summary = await leaveSpace();
// { duration: '300s', transcriptions: 42, responses: 8 }

MCP Tools:

📖 Full guide: docs/spaces-agent.md: configuration, environment variables, events, multi-agent setup, and examples.

💻 CLI Reference

56 commands, grouped by task, with 85 subcommands under them. Run xactions with no arguments for the grouped screen, xactions help <command> for one command, and xactions completion bash|zsh|fish for tab completion generated from the live command tree.

Start here quickstart doctor connect login logout mcp-config skills drafts info

xactions quickstart         # guided first run, adapts to what you already have
xactions doctor             # what works right now, and what each failure needs
xactions connect            # log in through a real browser, no DevTools
xactions login --from-browser chrome    # or read cookies from an installed browser
xactions login --cookies-file cookies.txt   # or import a cookie export
xactions skills install --all           # 50 agent skills into Claude Code, Cursor, Codex, Windsurf
xactions drafts list                    # MCP writes waiting for your approval

Read an account (no login needed) profile tweets thread media analyze report history snapshot

xactions profile NASA
xactions tweets NASA --limit 50 --output nasa.csv
xactions thread https://x.com/NASA/status/1234567890
xactions media NASA --limit 30
xactions analyze NASA SpaceX          # engagement rate, cadence, content mix, best hour

Followers and audience followers following non-followers audience crm graph

xactions followers YOUR_USERNAME --limit 1000 --output followers.json
xactions non-followers YOUR_USERNAME  # who does not follow back
xactions audience nasa spacex         # follower overlap between two accounts
xactions graph build nasa             # build and analyse a social graph

Search and monitor search hashtag scrape platforms monitor sentiment rss stream

xactions search "your brand" --filter latest --limit 50
xactions hashtag ai --limit 50
xactions monitor yourbrand            # sentiment over time
xactions stream start tweet nasa      # tweet, follower or mention streams

Write and grow engage ai optimize hashtags predict variations evergreen persona schedule bulk

xactions engage USERNAME --like --repost --comment --prompt "supportive, one honest question"
xactions ai generate "shipping fast" --voice nichxbt
xactions evergreen YOUR_USERNAME      # find and recycle top performers
xactions persona run my-persona

Automate workflow agent notify plugin team dataset

xactions workflow run daily-digest
xactions notify test slack

Move data export export-data archive import convert migrate diff

xactions export YOUR_USERNAME             # profile, tweets, followers, following, bookmarks
xactions archive summary ~/Downloads/twitter-archive.zip
xactions archive migrate ~/Downloads/twitter-archive.zip --to bluesky --execute
xactions diff exports/january exports/february

Low level client

xactions client profile nasa          # the raw HTTP client, no browser at all

Output flags for agents and pipes

Global on every read command:

xactions profile NASA --compact                              # one record per line, key=value
xactions tweets NASA --limit 200 --fields id,likes,text --compact
xactions tweets NASA --limit 200 --json | jq 'sort_by(-.likes) | .[:5]'

--compact prints tab-separated key=value pairs with no colours or spinners. --fields narrows that to the columns you name, in the order you name them. --json prints the full structured object and outranks --output, so a pipe is never silently turned into a file write. If both --compact and --json are passed, --compact wins.

Public reads work with no account: profile, tweets, thread, media, analyze, hashtag. Search, followers, following, non-followers, likes, bookmarks and DMs need a session. Full reference: docs/cli-reference.md.

📚 Node.js API

Quick Start

import { 
  createBrowser, 
  createPage, 
  loginWithCookie,
  scrapeProfile,
  scrapeFollowers,
  scrapeFollowing,
  scrapeTweets,
  searchTweets,
  exportToJSON,
  exportToCSV 
} from 'xactions';

// Initialize
const browser = await createBrowser({ headless: true });
const page = await createPage(browser);

// Optional: Login for private data
await loginWithCookie(page, 'your_auth_token_cookie');

// Scrape profile
const profile = await scrapeProfile(page, 'elonmusk');

// Scrape followers with progress
const followers = await scrapeFollowers(page, 'elonmusk', {
  limit: 1000,
  onProgress: ({ scraped, limit }) => console.log(`${scraped}/${limit}`)
});

// Export data
await exportToJSON(followers, 'followers.json');
await exportToCSV(followers, 'followers.csv');

await browser.close();

All Scraper Functions

// Profile
scrapeProfile(page, username)

// Followers & Following
scrapeFollowers(page, username, { limit, onProgress })
scrapeFollowing(page, username, { limit, onProgress })

// Tweets
scrapeTweets(page, username, { limit, includeReplies, onProgress })
searchTweets(page, query, { limit, filter: 'latest'|'top' })
scrapeThread(page, tweetUrl)
scrapeHashtag(page, hashtag, { limit, filter })

// Media
scrapeMedia(page, username, { limit })
scrapeLikes(page, tweetUrl, { limit })

// Lists
scrapeListMembers(page, listUrl, { limit })

// Export
exportToJSON(data, filename)
exportToCSV(data, filename)

🌐 Don't Want to Code?

🔒 Safety & Best Practices

[!CAUTION] Do: Use 2–5s delays · Mix automated with manual activity · Test with small batches

Don't: Mass-follow thousands/day · Run 24/7 · Spam comments

🛠️ Built With

📁 Project Structure

Every top-level directory, so nothing here is a surprise.

src/            The library. cli/ mcp/ scrapers/ client/ automation/ agents/
                analytics/ streaming/ workflows/ plugins/ portability/ graph/
                a2a/ spaces/ ai/ auth/ scheduler/ notifications/, plus the
                browser console scripts that sit at the top level
api/            Express REST API: routes/ services/ middleware/ realtime/
bin/            Legacy command name (unfollowx), forwards to the CLI
packages/       xactions-mcp, the thin MCP wrapper published to npm
skills/         50 Agent Skills, one directory each
extension/      Chrome and Edge extension (Manifest V3)
integrations/   Third-party glue (n8n and friends)
worker/         Cloudflare Worker entry point
functions/      Cloudflare Pages Functions for the /api/* edge surface

site/           The marketing landing page served at /
dashboard/      The signed-in app shell at /dashboard, plus the generated
                docs, guides, skills and script pages the site serves
public/         Static assets: OG images, robots.txt, sitemap
playground/     Standalone in-browser playground

docs/           Documentation source (Markdown)
tutorials/      Guided walkthroughs
examples/       Runnable programs, from a profile lookup to the MCP draft gate
prompts/        Prompt library for coding agents

tests/          Vitest suite
scripts/        The 95 browser console scripts, plus build, docs and
                maintenance tooling. twitter/ holds standalone console variants
config/         Agent, niche and persona configuration
prisma/         Database schema and migrations
deploy/         Cloudflare and GCP deployment assets
types/          TypeScript declarations
data/           Runtime output (gitignored, keeps a .gitkeep)
archive/        Retired code, kept for reference. Do not modify

python/         xeepy, the Python twin of the library
xspace-agents/  The Spaces agent monorepo

Both python/ and xspace-agents/ are self-contained projects that live in this repository but build on their own. Start from their READMEs, not this one.

🤝 Contributing

Contributions welcome. See CONTRIBUTING.md, and AGENTS.md for the repository map and the three runtime contexts.

git clone https://github.com/nirholas/xactions.git
cd xactions && npm install
npm run cli -- profile elonmusk   # run the CLI from the source tree
npm run mcp                       # run the MCP server

Node.js 20 or newer is required (engines.node is >=20); CI runs the suite on 20, 22 and 24. Three commands before you open a PR, all of which run in CI:

npm test              # the whole suite, offline, under a minute
npm run lint          # ESLint over the repo
npm run docs:check    # dead links, stale counts, invented CLI commands

⭐ Star History

If XActions saved you from paying $100/mo for Twitter's API, star the repo: it's how open source grows.

📚 Full Tutorial Library

XActions is 100% free and open source. Visit xactions.app for interactive tutorials.

🚀 One-Click Script Runner

NEW! Run scripts without any coding knowledge:

1. Visit xactions.app/run.html
2. Drag any blue button to your bookmarks bar
3. Go to x.com and click the bookmarklet

No console, no code, no setup!

Quick Links by Category

All Documentation

• Getting Started
• CLI Reference
• Automation Guide
• Monitoring Guide

Example Docs (Full Code)

View the full README on GitHub