x-use-mcp 2.2.0

Browser-native AI agents for X (Twitter): multi-account, MCP-ready, no API keys required.

x-use

Browser-native AI agents for X (Twitter). Multi-account, MCP-ready, no X API key required.

x-use drives a real, stealth-hardened browser instead of the paid X API. It posts, replies, searches, and engages across as many accounts as you configure, writes content with your own LLM, and exposes everything as MCP tools, so Claude Desktop, Claude Code, Cursor, and other MCP clients can run your X presence directly.

The X API's pricing tiers put write access out of reach for exactly the people who want to automate a couple of accounts. x-use sidesteps the API entirely: if a logged-in browser can do it, an MCP client can ask for it. And because write actions go through a draft-approval step by default, an agent can prepare work all day while nothing reaches X until a human says yes.

x-use is the v2 relaunch of twitter-automation-ai. The repository was renamed; old URLs keep redirecting, and stars, forks, and issues came along intact.

---

Install

From PyPI (CLI and MCP server):

pip install x-use-mcp

For the full repo setup (presets, example configs, docs), the one-line installer clones the repo, installs x-use into its own virtual environment, and finishes with x-use doctor so you can see what is left to configure.

Windows (PowerShell):

iex "& { $(irm https://raw.githubusercontent.com/ihuzaifashoukat/x-use/main/install.ps1) }"

macOS / Linux / Git Bash:

curl -fsSL https://raw.githubusercontent.com/ihuzaifashoukat/x-use/main/install.sh | bash

Or the manual way:

git clone https://github.com/ihuzaifashoukat/x-use.git
cd x-use
pip install -e .

Requires Python 3.10+ and Chrome. Any of these gives you the x-use command.

Set up

x-use init     # interactive wizard: presets, account + cookie import, LLM keys
x-use doctor   # verify browser/driver, cookies, LLM keys, proxies

Then connect your AI client. Paste this into claude_desktop_config.json (Claude Desktop > Settings > Developer > Edit Config); the same command/args pair works for any MCP client that runs stdio servers:

{
  "mcpServers": {
    "x-use": {
      "command": "x-use",
      "args": ["mcp"]
    }
  }
}

If x-use is not on your client's PATH, use the full path the installer printed (for example venv/bin/x-use or venv\Scripts\x-use.exe). Restart the client, then ask it to list_accounts.

Draft mode is on by default. Write tools return a reviewable draft and change nothing until you call approve_draft with the returned draft_id. Opt out with "mcp": { "draft_mode": false } in config/settings.json.

MCP tools

25 tools in four groups. Two safety gates: the immediate write tools run in draft mode by default (review, then approve_draft), and the queue only stores work until an explicit process_queue call (or the opt-in auto_drain worker) executes it with jittered pacing and daily caps.

Interactive use needs no LLM key at all: your MCP client (Claude, Codex, ...) does the thinking and passes explicit text to the write tools. The optional server-side LLM (llm block in settings) only powers "auto" generation and background automation.

Read-only and status:

Tool	What it does
list_accounts()	List configured accounts with secrets stripped.
get_account(account)	One account's masked config plus cookie-file status.
get_metrics(account)	Counters and recent events for an account.
search_tweets(keywords, limit=10, account?)	Search recent posts for a query.
prepare_reply(account, tweet_url)	Fetch a tweet's content plus the account context so your agent can write the reply itself. No LLM, nothing posted.
list_queue(account?, status?)	Queued actions with full payloads (exactly what will fire) plus per-status counts.
list_drafts(status?, account?, limit=20)	Drafts, newest first, with filters.
get_draft(draft_id)	One draft by id.
reject_draft(draft_id)	Reject a pending draft (local status flip; nothing touches X).
get_run_status(run_id?)	Poll run_cycle handles.
get_account_health(account)	Config, cookie validity, metrics, session, queue, and draft state in one call.

Write tools (draft-gated):

Tool	What it does
post_tweet(account, text, media?, community?)	Post text/media, optionally into a community.
generate_and_post(account, topic)	Generate a post with the server-side LLM (needs the llm block), then post it. Prefer post_tweet with your own text when driving from an MCP client.
reply_to_tweet(account, tweet_url, text="auto")	Reply with explicit text, or "auto" to generate from the tweet's content via the server-side LLM.
engage(account, keywords, actions=["like"], max_actions=5)	Relevance-gated likes/retweets on keyword results.
run_cycle(account?, pipelines?)	Full orchestrator cycle in the background; returns a run handle.
approve_draft(draft_id)	Execute a pending draft exactly once.

Scheduled queue (the second gate):

Tool	What it does
queue_post(account, text? topic?, media?, community?, not_before?)	Queue a post for paced execution. topic generates the text now, so list_queue shows the final payload. not_before schedules (ISO 8601).
queue_engagement(account, action, tweet_url, text?)	Queue a like, retweet, or reply. Replies accept "auto" to generate the text now.
cancel_queued_action(queue_id)	Cancel a pending or failed item.
process_queue(account?, max_actions=5)	The approval gate: drains due items through the shared pacing/dedup/metrics path with daily caps.

Account management (validated, backed up, atomic):

Tool	What it does
add_account(account_id, cookie_file?, proxy?, target_keywords?, is_active=true)	Add an account. Cookies import from a file path on the server, never inline.
update_account(account, ...)	Partial update; closes the warm session so changes take effect.
set_account_active(account, active)	Pause or resume an account without deleting it.
remove_account(account, confirm)	Remove an account (requires confirm=true; backups in config/backups/).

Drafts persist in data/drafts.jsonl; the queue persists in data/engagement_queue.jsonl. Both survive restarts.

CLI

x-use init                            # interactive setup wizard
x-use run                             # all active accounts, concurrent
x-use run --account my_account        # one account only
x-use run --pipeline keyword_replies  # one pipeline only (in-memory; config files untouched)
x-use doctor                          # environment checks; exits non-zero on failure
x-use mcp                             # start the MCP stdio server

Pipelines for --pipeline: community_engagement, competitor_reposts, content_curation, keyword_replies, keyword_retweets, likes. The MCP run_cycle tool accepts the same names.

The legacy python src/main.py entry point still works via a deprecation shim (removal no earlier than v2.1).

Features

Area	What you get
MCP server	25 tools over stdio on the official MCP Python SDK (FastMCP, pinned mcp>=1.6,<2): draft-gated writes, a persistent scheduled-action queue with daily caps, account management, and a lazy per-account browser session pool.
Draft mode	On by default. Write tools build the full payload (including LLM-generated text), store a draft, and touch nothing until approve_draft runs.
Multi-account engine	Post (including communities and media), reply, repost/quote, like, keyword search, and relevance-gated engagement. Per-account overrides for keywords, LLM settings, and action behavior.
LLM generation	One OpenAI-compatible client (llm: api_key, base_url, model) covers OpenAI, OpenRouter, Azure, Gemini, and local servers. Only needed for "auto" text and background automation; interactive MCP use runs keyless. Keys resolve from env/.env first, then config/settings.json.
Stealth	undetected-chromedriver, selenium-stealth, randomized user agents, headless support.
Proxies	Per-account proxy, named pools, hash or round-robin rotation, ${VAR} env interpolation in proxy strings.
Metrics	Per-account counters in data/metrics/<account_id>.json plus JSONL event logs in logs/accounts/<account_id>.jsonl.

x-use vs. API-based X MCP servers

	x-use	API-based X/Twitter MCP servers
X API cost	$0: cookie auth, no X API key needed	Paid X API tier required
Multi-account	Built-in: per-account config, cookies, proxies	Typically one account
Proxies	Per-account proxies, named pools, hash/round-robin rotation	N/A
Stealth	undetected-chromedriver + selenium-stealth, randomized user agents	N/A (official API)
Write safety	Draft mode on by default, explicit approve_draft gate	Usually posts directly
Metrics	Per-account counters + JSONL event logs, readable via MCP	Varies

An LLM key is optional: interactive MCP use needs none (your agent writes the text). For "auto" generation and background automation, set one OpenAI-compatible key (llm.api_key + base_url + model, or OPENAI_API_KEY/OPENAI_BASE_URL/OPENAI_MODEL); that is the only key involved.

Configuration

• config/accounts.json: your accounts (gitignored). Start from config/accounts.example.json or let x-use init write it.
• config/settings.json: global defaults for browser, pacing, action caps, LLM, proxies, and the mcp section.
• .env: LLM API keys; overrides settings.json (env wins). See .env.example.

Full schema: docs/CONFIG_REFERENCE.md. Starter templates: presets/ (offered as wizard choices by x-use init).

Recommended proxy providers

X's per-IP detection kills automation at scale. A quality proxy provider keeps a multi-account setup running reliably. The providers below are tested and recommended.

ScrapingAnt (featured)

Private residential network with premium-grade IPs, not resold traffic. Supports the sticky sessions that X's IP consistency checks require.

Top-notch, fast residential ScrapingAnt’s proxies for best performance

Affordable datacenter proxies for cost-effective operations

> 5% discount for x-use users: apply code `TWI_AUTO` at checkout.

RapidProxy

RapidProxy

Dynamic and static residential proxies with free testing and unlimited, non-expiring traffic.

Responsible use

Browser automation of X carries real account risk. Read BEST_PRACTICES.md before running anything: it covers conservative rate limits (the shipped defaults), account warm-up, relevance filters, cookie and credential hygiene, and X ToS considerations. Keep delays high, caps low, and draft mode on.

Development

pip install -e '.[dev]'
pytest

252 tests cover config loading and merging, dedup keys, LLM JSON extraction and the single-client service, tweet parsing, proxy pool selection, the MCP tool contract, drafts, sessions, the action queue, account writes, and the CLI; none of them needs a network or a browser. CI (.github/workflows/ci.yml) runs the suite plus an import smoke check on Python 3.10/3.11/3.12.

The v2.0 relaunch is merged; PyPI publishing and MCP directory submissions are the remaining Phase 1 items. Dashboard and Docker are Phase 2; personas, plugins, and selector self-healing are Phase 3. See ROADMAP.md.

Contributing

Contributions are welcome. Selector fixes, presets, docs, and MCP tool ideas (via issues) are the most valuable contributions right now. See CONTRIBUTING.md for the workflow and CODE_OF_CONDUCT.md for community expectations.

Star history

License

MIT. See LICENSE.