Bux Claude Agent logo

Bux Claude Agent

Organization
reason-machines
bux-claude-agent

Deploy a 24/7 Claude Code agent with persistent Chromium browser on any Ubuntu VPS, controllable via Telegram

Overview

Publisherreason-machines
Repositorytrending-skills
Skill namebux-claude-agent
Stars
89
Forks
16
Bundled files
Instructions only
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by reason-machines on GitHub. Read the source before you install it.

Installation

Install the Bux Claude Agent AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/reason-machines/trending-skills/tree/main/skills/bux-claude-agent
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/reason-machines/trending-skills.git /tmp/trending-skills
mkdir -p .claude/skills
cp -r /tmp/trending-skills/skills/bux-claude-agent .claude/skills/bux-claude-agent
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Bux Claude Agent in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Bux Claude Agent on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Bux Claude Agent is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Bux (Browser Use Box) ♞

Skill by ara.so — Daily 2026 Skills collection.

Bux turns any Ubuntu VPS into a 24/7 Claude Code agent with a persistent Chromium browser session. Text your agent via Telegram, it browses the web, logs into sites, handles 2FA interactively, and maintains state across reboots — all on hardware you own.

What Bux Does

  • Runs Claude Code as a persistent systemd service on your server
  • Keeps a real Chromium browser alive (cookies/logins persist across sessions)
  • Exposes a Telegram bot interface so you can message your agent from anywhere
  • Provides a web terminal (ttyd) bound to localhost for direct access
  • Hands you a live browser view URL when human interaction is needed (CAPTCHA, 2FA, login walls)
  • All state lives in /home/bux — survives reboots

Requirements

  • Ubuntu 22.04+ VPS with ≥ 2GB RAM (a $5/mo Hetzner or DigitalOcean droplet works)
  • Browser Use Cloud API key — BROWSER_USE_API_KEY=bu_xxx
  • Anthropic API key or Claude Max subscription (entered during /login)
  • (Optional) Telegram bot token from @BotFather

Installation

One-liner install

bash
curl -fsSL https://raw.githubusercontent.com/browser-use/bux/main/install.sh \
    | sudo BROWSER_USE_API_KEY=bu_xxx bash

With Telegram bot enabled

bash
curl -fsSL https://raw.githubusercontent.com/browser-use/bux/main/install.sh \
    | sudo BROWSER_USE_API_KEY=$BROWSER_USE_API_KEY \
           TG_BOT_TOKEN=$TG_BOT_TOKEN bash

Via Claude Code (automated setup)

Paste this prompt into Claude Code on your laptop:

Set up https://github.com/browser-use/bux on my remote box.

SSH into it (host: YOUR_SERVER_IP), run install.sh with my BROWSER_USE_API_KEY,
and wire up a Telegram bot with my TG_BOT_TOKEN. Read install.md first.
After the install completes, verify the services are running
(systemctl is-active bux-browser-keeper bux-ttyd), then become the `bux`
user and run `claude /login` so I can complete the OAuth flow.
Once logged in, test by asking claude to visit https://browser-use.com
and report the page title.

Architecture

  telegram ──►  telegram_bot.py ─┐
                                 ├──► claude -p  ──► browser-harness ──► BU Cloud
  browser  ──►  ttyd ────────────┘         │            (cdp over wss)
                                           ▼
                                  /home/bux (persistent state)

Three systemd services:

ServicePurpose
bux-browser-keeperKeeps Chromium alive via browser-harness
bux-ttydWeb terminal on localhost
bux-telegramTelegram bot bridge (if token provided)

Verifying the Install

bash
# Check all services are active
systemctl is-active bux-browser-keeper bux-ttyd bux-telegram

# View logs for each service
journalctl -u bux-browser-keeper -f
journalctl -u bux-telegram -f
journalctl -u bux-ttyd -f

# Check as the bux user
sudo -u bux bash
cd ~
ls -la  # should show .claude, browser-harness state, etc.

Completing Claude Login

After install, you must log Claude in once:

bash
sudo -u bux claude /login
# Follow the OAuth flow printed to stdout

Key Configuration Files

/home/bux/agent/CLAUDE.md

Context loaded by Claude on every session. Edit this to give your agent persistent instructions, personas, or domain knowledge.

markdown
# My Agent Instructions

## Preferred behavior
- Always summarize email in bullet points
- When booking anything, confirm with me first
- Use my calendar at cal.example.com for scheduling

## Credentials locations
- Passwords are in ~/secrets/ (never log these)

Environment variables (set in systemd unit or shell)

bash
BROWSER_USE_API_KEY=bu_xxx        # Required: Browser Use Cloud key
TG_BOT_TOKEN=xxx                  # Optional: Telegram bot token
TG_ALLOWED_USERS=123456,789012    # Optional: comma-separated Telegram user IDs
ANTHROPIC_API_KEY=sk-ant-xxx      # Optional if using Claude Max via /login

Using Your Agent via Telegram

Once set up, message your bot directly:

You: check my gmail and summarize unread emails
Agent: [browses Gmail, returns summary]

You: book a flight from SFO to NYC next Tuesday under $300
Agent: [opens Kayak/Google Flights, searches, returns options]
       I found 3 options. Should I proceed with booking? [live view URL]

You: yes, book the Delta flight at 8am
Agent: [navigates to checkout] I need your payment info — 
       here's a live view: https://view.browser-use.com/session/xxx
       Please complete the payment form directly.

Working with the Browser Directly

When Claude hits an interactive wall, it surfaces a live view URL:

Claude: I've hit a 2FA prompt on your bank login.
        Live view: https://view.browser-use.com/session/abc123
        Please enter your 2FA code there, then tell me when done.

You open the URL in your browser, complete the action, then reply "done" — Claude continues automatically.

Deploying to Specific Providers

Hetzner Cloud

bash
# Create CX21 (2 vCPU, 4GB RAM) — plenty for bux
hcloud server create \
  --name my-agent \
  --type cx21 \
  --image ubuntu-22.04 \
  --ssh-key ~/.ssh/id_rsa.pub

# Get the IP
hcloud server ip my-agent

# Install bux
ssh root@<IP> "curl -fsSL https://raw.githubusercontent.com/browser-use/bux/main/install.sh \
  | sudo BROWSER_USE_API_KEY=$BROWSER_USE_API_KEY TG_BOT_TOKEN=$TG_BOT_TOKEN bash"

DigitalOcean

bash
# Create a $6/mo Basic Droplet via doctl
doctl compute droplet create my-agent \
  --size s-1vcpu-2gb \
  --image ubuntu-22-04-x64 \
  --region nyc1 \
  --ssh-keys $(doctl compute ssh-key list --format ID --no-header)

Raspberry Pi (local)

bash
# Ensure Ubuntu 22.04+ on Pi, then same install script
ssh pi@raspberrypi.local
curl -fsSL https://raw.githubusercontent.com/browser-use/bux/main/install.sh \
    | sudo BROWSER_USE_API_KEY=$BROWSER_USE_API_KEY bash

Customizing Agent Behavior

Give your agent persistent skills

Edit /home/bux/agent/CLAUDE.md as bux user:

bash
sudo -u bux nano /home/bux/agent/CLAUDE.md

Send one-off tasks programmatically

python
import subprocess

result = subprocess.run(
    ["sudo", "-u", "bux", "claude", "-p",
     "Visit https://news.ycombinator.com and return the top 5 headlines as JSON"],
    capture_output=True, text=True
)
print(result.stdout)

Invoke claude non-interactively from a cron job

bash
# /etc/cron.d/bux-morning-brief
0 8 * * * bux /usr/local/bin/claude -p \
  "Check my email, weather for San Francisco, and top HN stories. \
   Send a morning brief to my Telegram." \
  >> /var/log/bux-morning.log 2>&1

Telegram Bot Python Bridge (telegram_bot.py)

The bot bridges Telegram messages to claude -p. Simplified structure:

python
import asyncio
import subprocess
from telegram import Update
from telegram.ext import ApplicationBuilder, MessageHandler, filters, ContextTypes

ALLOWED_USERS = {int(uid) for uid in os.environ.get("TG_ALLOWED_USERS", "").split(",") if uid}

async def handle_message(update: Update, context: ContextTypes.DEFAULT_TYPE):
    user_id = update.effective_user.id
    if ALLOWED_USERS and user_id not in ALLOWED_USERS:
        await update.message.reply_text("Unauthorized.")
        return

    task = update.message.text
    await update.message.reply_text("On it...")

    result = await asyncio.to_thread(
        subprocess.run,
        ["claude", "-p", task],
        capture_output=True, text=True,
        cwd="/home/bux"
    )
    await update.message.reply_text(result.stdout or result.stderr)

app = ApplicationBuilder().token(os.environ["TG_BOT_TOKEN"]).build()
app.add_handler(MessageHandler(filters.TEXT & ~filters.COMMAND, handle_message))
app.run_polling()

Troubleshooting

Services not starting

bash
# Check status
systemctl status bux-browser-keeper
journalctl -u bux-browser-keeper --since "5 min ago"

# Restart all bux services
systemctl restart bux-browser-keeper bux-ttyd bux-telegram

# Verify bux user exists
id bux

Claude not authenticated

bash
sudo -u bux claude /login
# or check existing auth
sudo -u bux claude /status

Browser session lost

bash
# Restart the browser keeper
systemctl restart bux-browser-keeper
# Wait ~10s for Chromium to initialize
sleep 10
systemctl status bux-browser-keeper

Telegram bot not responding

bash
# Verify token is set
systemctl show bux-telegram | grep Environment

# Test the token directly
curl "https://api.telegram.org/bot${TG_BOT_TOKEN}/getMe"

# Check if your Telegram user ID is in the allowlist
# Get your ID by messaging @userinfobot on Telegram

Out of memory (2GB VPS)

bash
# Check memory usage
free -h
# Chromium + Claude can be memory-hungry; consider 4GB RAM droplets
# Or add swap:
fallocate -l 2G /swapfile
chmod 600 /swapfile
mkswap /swapfile
swapon /swapfile
echo '/swapfile none swap sw 0 0' >> /etc/fstab

Reinstalling / updating

bash
# Re-run the install script — it's idempotent
curl -fsSL https://raw.githubusercontent.com/browser-use/bux/main/install.sh \
    | sudo BROWSER_USE_API_KEY=$BROWSER_USE_API_KEY bash

Managed Alternative

If you don't want to manage a VPS: cloud.browser-use.com provisions the same stack for you in ~60 seconds with a one-command bux up Claude Code skill.

Contributing

The highest-leverage contribution is a deploy recipe — one .md file in docs/recipes/ documenting provider-specific setup steps (Oracle Free Tier, Fly.io, Mac mini, etc.).

bash
git clone https://github.com/browser-use/bux
# Add docs/recipes/YOUR_PROVIDER.md
# Submit a PR

Frequently asked questions

What does the Bux Claude Agent AI skill do?

Deploy a 24/7 Claude Code agent with persistent Chromium browser on any Ubuntu VPS, controllable via Telegram

Why use Bux Claude Agent on TypingMind?

Because you install it once and use it with any model. Bux Claude Agent is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Bux Claude Agent in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/reason-machines/trending-skills/tree/main/skills/bux-claude-agent. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Bux Claude Agent?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Bux Claude Agent?

As many as you like. As long as a model supports skills, you can use Bux Claude Agent with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Bux Claude Agent AI skill free?

It is published on GitHub by reason-machines. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇