Your browser. Your voice. Your X growth, on autopilot.
A terminal agent that audits your X account and drafts on-voice replies to the right people, driven by your real browser session and your own LLM key. Read-only and draft-and-approve by default.
cancel anytime · bring your own browser + LLM
your llm drives it
$ claude mcp add tweets-bot -- node ./src/mcp.js✓ tweets-bot connected · 6 tools› audit my account, find fresh posts from my targets, draft on-voice replies. show me before posting.⛁ audit_account(devhanzo1)✓ voice profile saved · niche: indie hacking⛁ find_targets(from: targets, since: 60m)✓ 10 fresh, relevant targets⛁ draft_replies()✓ 10 on-voice drafts · nothing posted yetDrafts ready to review. Approve and I'll post them: paced, capped, your call.›
how it works
Connect your own browser
Launch your logged-in browser with a debug port. tweets-bot drives that real session over CDP. It never logs in for you and never sees your X password.
Audit & learn your voice
It reads your profile and recent posts (read-only) with your own LLM key, then builds a reusable voice profile: tone, cadence, what's working.
Find the right people
Discover recent, relevant posts worth replying to, filtered by topic, recency, and engagement, ranked against your voice.
Draft, review, then post
On-voice replies land in a queue you edit. Nothing posts until you say so: paced, jittered, and capped under a hard daily limit.
why it's safer
Nothing is mimicked. It's genuinely your account.
Your real session, not a synthetic one
It drives your own browser with your own logged-in cookies. There's no proxy account and no mimicry. The fingerprint is genuinely yours because it is you.
Draft-and-approve by default
The default is read-only and draft-only. You click post. That keeps the action formally yours, not a bot's, and keeps you in control.
Paced, jittered, hard-capped
Randomized delays, an active-hours window, and a hard daily cap. The same careful pacing philosophy battle-tested in real reply runs.
Bring your own LLM
Plug in your Anthropic, OpenAI, or local Ollama key. Your prompts and your data go to your model, not through a black box.
pricing
Pro
cancel anytime · 3-day money-back
$100$49/mo
Growth on auto-pilot
The whole loop, driven by your own LLM. Bring your own browser and model key.
- Account audit + voice profile
- Guided setup for brand-new accounts
- Target discovery + your target list
- Draft, review, and safe posting (paced + capped)
- Reply to your own posts (inbox)
- MCP server for your LLM
- Priority selector updates
faq
Will this get my account banned?
No tool that automates engagement can promise zero risk, and we won't pretend otherwise. What we can say plainly: it runs in your own real browser session, defaults to draft-and-approve so you click post, and enforces human-like pacing under a hard daily cap. That's the safest posture in this category.
Do you see my X password or session?
Never. You stay logged in to your own browser; tweets-bot attaches to it over a local debug port. We don't log in for you and we don't store credentials.
Whose LLM does it use?
Yours. Bring an Anthropic, OpenAI, or local Ollama key. You choose the model and pay your own inference, so your data doesn't route through us.
Do I need Claude Code or another AI agent?
It's the best experience but not required. tweets-bot runs as an MCP server, so an agent like Claude Code, Claude Desktop, or Cursor can drive the whole loop from one natural-language prompt and pause for your approval. If you'd rather not, every step is also a plain terminal command you run yourself.
I'm brand new to X. Does this still work?
Yes, and it starts somewhere different. With no posts there's no voice to learn yet, so setup detects a new account and offers a guided profile builder: a short interview becomes your positioning, a few bio options, and what to pin. It prefills your bio on the edit screen (you click save), and writes a starter voice profile so drafting works from day one.
Is it ToS-compliant?
Automated engagement isn't covered by X's official API, so nobody honest in this space claims full ToS-compliance. We minimize risk instead: real session, your approval on every post, and conservative limits.
Can I cancel? How do refunds work?
Cancel anytime from your account and you won't be billed again. First-time subscribers can also request a full refund within 3 days of the first payment (one per customer, returned to the original payment method). Full details on the terms page.