trackmcp
Back to directory

⚡ The Complete X/Twitter Automation Toolkit — Scrapers, MCP server for AI agents (Claude/GPT), CLI, browser scripts. No API fees. Open source. Unfollow people who don't follow back. Monitor real-time analytics. Auto follow, like, comment, scrape, without API. Follow Bot. Like bot. Grow your account automatically.

506 stars HTMLOthers Updated Sep 4, 2026
ai-agentautomationclaudegptmass-unfollowmcpmodel-context-protocolnodejsopen-sourcepuppeteertwittertwitter-automationtwitter-bottwitter-scraperunfollow-toolxxactionsmcp-server

Documentation


bash
npx xactions profile nasa
code
⚡ @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.

bash
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

🏆 How XActions Compares

> Why build with XActions instead of the alternatives?

FeatureXActionstwikittwscrapex-usexmcp (official)birdtwitter-cliAgent-Reach
No API Key Required✅ Cookies✅ Account pool✅ Browser❌ X API keys✅ Cookies❌ X API keys
MCP Server (AI agents)153 tools✅ (33-tool server)✅ (140-tool, metered)❌ Skills, not MCP
Browser Console Scripts✅ 95
AI Voice Agent in Spaces✅ Join, listen, speak
CLI✅ 56 commands
Human approval gate on writes✅ Every write held as a draft
Daily action caps that survive a restart✅ Per account, on disk
Account pool with rotation✅ SQLite, per-operation windows
Resumable long scrapes✅ Cursor checkpoints
Cookies from your installed browser✅ `--from-browser`, `--cookies-file`
Official X archive import✅ `xactions archive`
LanguageJavaScriptPythonPythonTypeScriptTypeScriptTypeScriptPythonPython

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

AreaWhat changed
Delete every reply you sent one account`scripts/searchSweep.js` turns any X search into a bulk action. Open `x.com/search?q=from:you @someone&f=live`, paste it, and delete, like, repost, or reply to every result, with a floating panel, a dry run, filters that protect posts that did numbers, and multiple passes because X search returns a slice at a time. Deletes only ever touch your own posts. Docs.
Sweep a whole profile`scripts/engageProfile.js` likes, reposts, and replies to every post on an account from the console, with a floating panel, dry run, resume, undo, and replies from your templates or from an LLM given a one-line brief (Grok works straight from the console; any provider through the extension). `xactions engage USERNAME --like --repost --comment --prompt "..."` does the same from the terminal with any provider. Docs.
A CLI you can find your way aroundAll 56 commands are now grouped by task instead of listed alphabetically. `xactions quickstart` gives a guided first run that adapts to what you already have set up.
Tab completion`xactions completion bash\zsh\fish` prints a completion script generated from the live command tree, so every command, sub-command, and flag completes.
Log in without touching DevTools`xactions login --from-browser` reads x.com cookies straight out of an installed Chrome, Chromium, Brave, Edge, Arc or Firefox profile. `--cookies-file` imports a Netscape `cookies.txt`, a Cookie-Editor or EditThisCookie JSON export, or a Playwright / Puppeteer `storageState`. A full cookie jar beats a bare `auth_token`, because it carries the `ct0` every write needs.
Query IDs that heal themselvesX rotates its GraphQL query IDs and every pinned client 404s at once. Ours are discovered from x.com's own bundles and cached, so a rotation is invisible. `xactions doctor` reports the cache age.
Requests signed like the real clientEvery GraphQL call carries an `x-client-transaction-id` header computed the way x.com's own web client computes it, so reads get the answer the browser gets.
An account pool, and scrapes that resumeSessions live in a SQLite database with their own proxy and a per-operation rate-limit window read from X's own headers. The pooled client rotates on a 429 and locks an account on a 401. A checkpoint written after every page means a 50,000-follower scrape that dies at page 400 restarts from its cursor, not from page one.
Daily action capsEvery MCP write is charged against a rolling 24 hour per-account budget stored on disk. The cap survives a restart, a crash, and a fresh `npx xactions-mcp`, and a call that would go over is refused before it reaches X. Defaults follow X's own published limits.
Writes a human releases`XACTIONS_MCP_REQUIRE_APPROVAL=1` turns every write tool into a draft. Review with `xactions drafts list`, release with `xactions drafts approve `, or use the `x_list_drafts` / `x_approve_draft` / `x_discard_draft` tools.
MCP tool groupsThe tool list is filterable: `--tools read,analytics` or `XACTIONS_MCP_TOOLS=read` advertises only what the session needs, and a filtered tool is neither advertised nor callable. Groups: read, write, dm, lists, spaces, analytics, ai, grok, automation, monitoring, workflows, persona, graph, data, x402, drafts, auth.
MCP over HTTP`xactions-mcp --http` serves the Streamable HTTP transport on `/mcp` for remote and hosted clients, with optional bearer auth via `XACTIONS_MCP_TOKEN`. stdio is still the default.
One-drag install for Claude DesktopThe `.mcpb` bundle carries the server and its dependencies and prompts for the session cookie and tool groups at install time, so nothing is typed into a config file. Attached to every release.
Skills you can install`xactions skills install --all` copies the 50 bundled skills into Claude Code, Cursor, Codex, Windsurf, or the current project. `xactions skills show ` reads one without installing it.
Bring your X archive`xactions archive summary\export\migrate` reads the official X data export zip: counts, date range, busiest year and top hashtags; a rewrite to JSON, CSV, Markdown and a browsable HTML viewer; or a migration to Bluesky or Mastodon. Docs.
Live event streaming`src/streaming/livePipeline.js` reads x.com's own event pipeline: engagement counters, DM updates and typing indicators, pushed rather than polled, with subscriptions changed mid-session. It is a chunked newline-delimited JSON response, not a WebSocket, because x.com never answers that endpoint with a 101.
Signed webhooksOutbound deliveries carry an HMAC-SHA256 signature, a timestamp, an event type and a stable delivery id, retry three times with backoff, and land in a replayable delivery log. Receivers verify with the exported `verifyWebhookSignature`.
`--compact` and `--fields`Global on every read command. `--compact` prints one record per line as tab-separated `key=value` pairs with no colours or spinners; `--fields id,text,likes` narrows that to the columns you name. Built for agents and pipes.
`--json` everywhereEvery read command accepts `--json` and puts data on stdout and nothing else. It outranks `--output`, so a pipe is never silently turned into a file write.
ReliabilityPublic reads no longer need a browser at all. Profiles and timelines come from the internal GraphQL API in about a second, with no Chromium download.
Honest failuresNothing reports empty results as success anymore. A read that cannot complete says why, and says what to do about it.
Archive every photo and video`xactions download @nichxbt:all --archive` saves a profile's media, avatar and banner with gallery-dl-style filename templates, then re-runs incrementally. Resumes interrupted files with `Range`, retries rate limits, and hard-links the same photo reached through a retweet instead of storing it twice. Also `xactions/media` and the `x_download_media` MCP tool. Docs.
Ask XActionsxactions.app/ask answers "how do I..." in plain language from these docs, the skills, the scripts and the repo, and ends every answer with the thing to run: the console script (copied from source), the terminal command, or the MCP tool. Also `xactions ask "how do I unfollow everyone?"` in a terminal, and `x_ask` so an agent can read the manual before it acts. Free LLM lanes, three of which need no key. Docs.
Examples`examples/`: 9 runnable programs, from a first profile lookup to the MCP draft-approval gate. Each one runs as written.
Tutorials`tutorials/`: six guided walkthroughs, from a first scrape with no account to scripting XActions inside your own pipelines.
Docs that stay true`npm run docs:check` fails the build on a dead link, a stale version, a wrong tool count, or a documented command that does not exist.
Cross-platformBluesky, Mastodon, and Threads scrapers behind one normalised interface.
Plugin systemCommunity plugins via npm: `xactions-plugin-*` packages extend scrapers, MCP tools, and automation.
Workflow engineDeclarative JSON pipelines with triggers, conditions, and chained actions.
Sentiment analysisOffline rule-based analyzer, with an optional LLM mode behind the same interface.
Browser extensionChrome/Edge Manifest V3: run automations from a popup, no console needed.

> [!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

bash
npm install xactions

CLI

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

Browser

No install needed. Copy-paste scripts into your browser console on x.com


📖 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 It Works

1. PasteCopy a script into thex.com DevTools console

2. RunUse the CLI, MCP server,or browser extension

3. DoneEverything runs locally.No data leaves your machine

> 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.

🤖 Optional: Remote API for AI Agents

If you self-host the XActions API for remote AI agent access, you can optionally enable pay-per-request micropayments via the x402 protocol. This is entirely optional and disabled by default.

OperationPrice
Profile scrape$0.001
Followers/Following$0.01
Tweet scrape$0.005
Search tweets$0.01
Unfollow non-followers$0.05
Detect unfollowers$0.02
Auto-like$0.02
Video download$0.005

This only applies to the hosted remote API. Local mode is always free.


🎯 Why XActions?

XActions

Others

ScopeBrowser + CLI + Node.js + MCP + ExtensionUsually 1 thing

API KeyNot neededMost need Twitter API ($100/mo)

MCP Tools152 for Claude, GPT, Cursor0-2 tools

AI FeaturesSentiment, Grok, reputationNone

ExportJSON, CSV, Markdown, HTMLJSON only (if any)

MigrationBluesky and Mastodon, from a live account or an X archive zipNone

Tutorials23 Claude promptsNone


🐳 Docker

Run XActions anywhere with one command:

bash
# 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:

bash
docker compose up

See Dockerfile for details.


📖 API Reference

Full TypeScript-compatible API with type declarations included.

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

Core Functions:

FunctionDescriptionReturns
`createBrowser(options?)`Launch Puppeteer browser`Browser`
`createPage(browser)`Create stealth page`Page`
`scrapeProfile(page, username)`Get user profile data`Profile`
`scrapeFollowers(page, username, options?)`List followers`User[]`
`scrapeFollowing(page, username, options?)`List following`User[]`
`scrapeTweets(page, username, options?)`Get user tweets`Tweet[]`
`searchTweets(page, query, options?)`Search tweets`Tweet[]`
`downloadVideo(page, tweetUrl)`Extract video URLs`VideoResult`
`exportBookmarks(page, options?)`Export bookmarks`Bookmark[]`
`unrollThread(page, tweetUrl)`Unroll a thread`Thread`

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.

TutorialWhat You'll Learn
MCP SetupInstall and connect XActions to Claude Desktop
Unfollow CleanupRemove non-followers, detect unfollowers
Growth SuiteAuto-follow, auto-like, keyword targeting
ScrapingExtract profiles, tweets, hashtags
Content PostingTweets, threads, polls, scheduling
AnalyticsPerformance tracking, competitor analysis
Autonomous Space AgentDeploy AI voice agents in live X Spaces
Power User Playbook10 advanced multi-feature strategies

**See all 23 tutorials →**


How to unfollow who doesn't follow back on X?

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

js
// 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();
})();

Or use the CLI or MCP server for more options.

How do I like, repost, and reply to every post on a profile?

1. Go to `https://x.com/USERNAME`

2. Open the Developer Console. (COMMAND+ALT+I on Mac)

3. Paste `scripts/engageProfile.js` and run it. A panel appears: pick like / repost / reply, leave dry run on, press Start, read the log, then turn dry run off.

Replies come from your own templates or from an LLM given a brief like "supportive, specific, one honest question". From the terminal: `npx xactions engage USERNAME --like --repost --comment --prompt "..."`. Full guide: docs/engage.md.

How do I mass unfollow on Twitter/X?

1. Go to `https://x.com/YOUR_USER_NAME/following`

2. Open the Developer Console. (COMMAND+ALT+I on Mac)

3. Paste the script into the Developer Console and run it

js
// 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();
})();

This script:

  • Is completely free.
  • Doesn't try and get you to sign in or take your personal data.
  • Automates your web browser to make it click unfollow buttons, scroll down to reveal more, then do it again.
  • No tricks, all of the code is here so you can see exactly what it does.

How do I find who unfollowed me on Twitter?

Use `src/detectUnfollowers.js` - it saves a snapshot of your followers and compares on next run.

How do I download Twitter/X videos?

Use `src/scrapers/videoDownloader.js` - extracts MP4 URLs from any tweet.

Twitter API alternative that's free?

XActions uses browser automation instead of the API. No API keys needed, no rate limits, no $100/mo fee.

Hypefury / Tweethunter alternative?

XActions is open-source and completely free for humans. AI agents pay micropayments per request.


⚠️ 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.

bash
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:

bash
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!*

javascript
// 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:

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

Node.js:

javascript
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:

javascript
// 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:

bash
xactions profile elonmusk --json

Node.js:

javascript
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:

javascript
// 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:

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

Node.js:

javascript
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:

javascript
// 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  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:

bash
# 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:

javascript
// 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  !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:

javascript
// 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

FeatureConsole ScriptCLINode.jsWebsite
SCRAPING
Scrape Profile
Scrape Followers
Scrape Following
Scrape Tweets
Search Tweets
Scrape Thread
Scrape Hashtag
Scrape Media
Scrape List Members
Scrape Likes
UNFOLLOW
Unfollow Non-Followers
Unfollow Everyone
Smart Unfollow (after X days)⚠️
Unfollow with Logging
FOLLOW
Follow User
Keyword Follow⚠️
Follow Engagers⚠️
Follow Target's Followers⚠️
ENGAGEMENT
Like Tweet
Retweet
Auto-Liker⚠️
Auto-Commenter⚠️
Post Tweet
MONITORING
Detect Unfollowers
New Follower Alerts
Monitor Any Account
Continuous Monitoring⚠️
COMMUNITIES
Leave All Communities⚠️⚠️⚠️
SPACES
Discover Live Spaces
Scrape Space Metadata
AI Agent Joins Space
Agent Listens & Speaks
ADVANCED
Multi-Account✅ Pro
Link Scraper
Growth Suite✅ Pro
Customer Service Bot✅ Pro
MCP Server (AI Agents)
Export to CSV/JSON

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`):

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:

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

Available MCP Tools (152)

CategoryTools
Scraping`x_get_profile`, `x_get_followers`, `x_get_following`, `x_get_tweets`, `x_search_tweets`, `x_get_thread`, `x_download_video`, `x_get_replies`, `x_get_hashtag`, `x_get_likers`, `x_get_retweeters`, `x_get_media`, `x_get_mentions`, `x_get_quote_tweets`, `x_get_likes`, `x_get_recommendations`
Analysis`x_detect_unfollowers`, `x_analyze_sentiment`, `x_best_time_to_post`, `x_competitor_analysis`, `x_brand_monitor`, `x_audience_insights`, `x_engagement_report`, `x_crypto_analyze`
Actions`x_follow`, `x_unfollow`, `x_like`, `x_post_tweet`, `x_post_thread`, `x_reply`, `x_retweet`, `x_quote_tweet`, `x_bookmark`, `x_send_dm`, `x_create_poll`, `x_delete_tweet`
Automation`x_auto_follow`, `x_follow_engagers`, `x_unfollow_all`, `x_smart_unfollow`, `x_auto_comment`, `x_auto_retweet`, `x_auto_like`, `x_engage`, `x_unfollow_non_followers`
AI`x_analyze_voice`, `x_generate_tweet`, `x_summarize_thread`, `x_rewrite_tweet`, `x_detect_bots`, `x_find_influencers`, `x_smart_target`, `x_grok_analyze_image`
Monitoring`x_monitor_account`, `x_monitor_keyword`, `x_follower_alerts`, `x_track_engagement`, `x_monitor_reputation`, `x_stream_start`
Workflows`x_workflow_create`, `x_workflow_run`, `x_workflow_list`, `x_workflow_actions`
Persona`x_persona_create`, `x_persona_run`, `x_persona_edit`, `x_persona_list`, `x_persona_presets`
Portability`x_export_account`, `x_migrate_account`, `x_diff_exports`, `x_import_data`, `x_convert_format`
Spaces`x_get_spaces`, `x_scrape_space`, `x_space_join`, `x_space_leave`, `x_space_status`, `x_space_transcript`
Graph`x_graph_build`, `x_graph_analyze`, `x_graph_recommendations`, `x_graph_list`
Drafts`x_list_drafts`, `x_approve_draft`, `x_discard_draft`, `x_draft_status`

Every tool belongs to exactly one group, and the list is filterable, so a client

loads only what the session needs:

bash
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

bash
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 ` and

`xactions drafts discard `.

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

bash
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 ` 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

bash
npm install xactions xspace-agent

Set your credentials:

bash
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:

javascript
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:

ToolDescription
`x_space_join`Join a Space with an autonomous AI voice agent
`x_space_leave`Leave the active Space and get session summary
`x_space_status`Get agent status (duration, transcription/response counts)
`x_space_transcript`Get recent transcriptions from the active Space

📖 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 ` 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`

bash
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`

bash
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`

bash
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`

bash
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`

bash
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`

bash
xactions workflow run daily-digest
xactions notify test slack

Move data `export` `export-data` `archive` `import` `convert` `migrate` `diff`

bash
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`

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

Output flags for agents and pipes

Global on every read command:

bash
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

javascript
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

javascript
// 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

Rate Limiting: Built-in 1-3s delays, human-like scrolling, auto-pause on rate limits

Auth Token: `x.com` → DevTools (F12) → Application → Cookies → copy `auth_token`

> [!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.

code
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.

bash
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:

bash
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!

CategoryScriptsTutorial
UnfollowUnfollow Everyone, Non-Followers, Smart UnfollowTutorial
AutomationAuto-Liker, Auto-Commenter, Follow EngagersTutorial
ScrapingVideo Download, Followers, Tweets, HashtagsTutorial
MonitoringDetect Unfollowers, Track Accounts, AlertsTutorial
CommunitiesLeave All CommunitiesTutorial
AI/MCPClaude Desktop, GPT IntegrationTutorial

All Documentation

Example Docs (Full Code)

FeatureDocumentation
Unfollow Everyoneunfollow-everyone.md
Unfollow Non-Followersunfollow-non-followers.md
Detect Unfollowersdetect-unfollowers.md
Auto-Likerauto-liker.md
Auto-Commenterauto-commenter.md
Follow Engagersfollow-engagers.md
Video Downloadervideo-downloader.md
Followers Scrapingfollowers-scraping.md
Tweet Scrapingtweet-scraping.md
Leave Communitiesleave-all-communities.md
MCP Servermcp-server.md
Monitor Accountmonitor-account.md
New Follower Alertsnew-follower-alerts.md

⚡ XActions, the complete X/Twitter automation toolkit

100% Free & Open Source · Apache 2.0 License

·

·

 

 


🌐 Run the MCP server over HTTP

stdio is the default transport and is what a local client wants. For a remote

or hosted client, the same server speaks

MCP Streamable HTTP

on `/mcp`, with the same 153 tools and the same tool-group filtering:

bash
npx xactions-mcp --http --port 3000
# or: MCP_TRANSPORT=http PORT=3000 npx xactions-mcp

Require a token on it by setting `XACTIONS_MCP_TOKEN`; clients then send

`Authorization: Bearer `.

Connect from any MCP client

json
{
  "mcpServers": {
    "xactions": {
      "type": "http",
      "url": "http://localhost:3000/mcp"
    }
  }
}

Check it from the shell

bash
curl -X POST http://localhost:3000/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Deploying it somewhere public? Run it behind TLS with `XACTIONS_MCP_TOKEN` set,

and prefer a narrow tool set (`--tools read,analytics`) over the full list.

Dockerfile and docs/deployment.md cover

Railway, Fly.io, Docker and Cloudflare.


🏆 Full Comparison Matrix

> Extended feature-by-feature comparison with every alternative.

FeatureXActionstwikittwscrapex-usexmcp (official)birdtwitter-cliAgent-Reach
Node.js Library❌ Python❌ Python❌ Python❌ Python
Workflow Engine
Sentiment Analysis✅ Built-in
Real-Time Streaming✅ Polling streams + x.com's live event pipeline✅ live_pipeline✅ API only
Account Export/Migration✅ JSON/CSV/HTML, plus the official X archive zip
Dashboard (No-Code)
Grok AI Integration✅ Separate pkg
Docker Support
TypeScript Types
MCP tool allowlist✅ 17 groupsn/an/an/an/an/a
MCP over Streamable HTTP✅ `--http` on `/mcp`n/an/an/an/an/a
Signed outbound webhooks✅ HMAC-SHA256, retries, replay
Claude Tutorials✅ 23 prompts
Multi-platform (X + others)✅ Bluesky, Mastodon, Threads
CostFreeFreeFreeFreeFree + X API planFreeFree + X API planFree

Comparison reflects each project's public README at time of writing (August 2026); open an issue if a column is out of date.

Documentation

Full documentation site: https://nirholas.github.io/XActions/

Frequently asked questions

What is XActions?

XActions is ⚡ The Complete X/Twitter Automation Toolkit — Scrapers, MCP server for AI agents (Claude/GPT), CLI, browser scripts. No API fees. Open source. Unfollow people who don't follow back. Monitor real-time analytics. Auto follow, like, comment, scrape, without API. Follow Bot. Like bot. Grow your account automatically.

How do I install XActions?

Open the GitHub repository and follow its README. Most MCP servers are added to your client's MCP config, then called by your agent.

Is XActions open source?

Yes — it is hosted on GitHub at https://github.com/nirholas/XActions and has 506 stars.

Related MCP tools

Run your own MCP server? See who uses it and what to fix.

Measure it with TrackMCP