Live on Hostinger VPS

Mission Control

Your personal AI command center. Monitor Javis, track token usage, and manage your autonomous AI workforce — all in one place.

Bot Online
72 Skills Loaded
Claude + Llama
72.61.207.250

📡 Live Status

Auto-updated every 15 min via Llama · Last: 2026-08-06 18:15 UTC

Status update unavailable (HTTP Error 404: Not Found)
Bot
Inactive
Ollama
Running
Disk
48%
Memory
2.0Gi/7.8Gi
Skills
72
Agents
36
Scheduled
0
Recent Changes
  • search_mercedes.py
  • clean_fake_sent.py
  • autoresearch/autoresearch.py
  • autoresearch/candidate.py
  • backtesting/sort_and_push.py
  • backtesting/auto_retrain.py
  • backtesting/strategy_hunt_v3.py
  • backtesting/strategy_discovery.py

How It Works

Every message you send on Telegram goes through an agentic loop — Javis reads your request, picks the right skill, runs scripts on the VPS, and sends results back.

You (Telegram) → Javis Bot (VPS) → Claude API or Ollama
                                        ↓
                               Reads all 27 SKILL.md files
                               Executes scripts on VPS
                                        ↓
                               Sends result back to you
Bot Location
/root/claude-ai/execution/
Skills Location
/root/claude-ai/.claude/skills/
Uptime
24/7 via systemd
Access
Your Telegram only

👥 AI Team

Your autonomous AI workforce. You direct via Telegram — Javis orchestrates and delegates to specialist agents.

👤
Paul
CEO
🤖
Javis
Chief of Staff
24/7 VPS · 72 Skills · Claude + Llama
Specialist Agents
📊
accountant-agent
Bookkeeper
QuickBooks invoicing, overdue detection, payment reminders, AR aging reports
🤖
agents-orchestrator
Specialist
---
📧
andre-email-agent
Email Triage
Classifies Gmail threads into Action Required, Waiting On, and Reference
🤖
btc-backtest-agent
Specialist
You are **Atlas** — an autonomous quantitative backtesting agent. Your job is to find profitable BTC
🤖
code-analyzer
Specialist
---
🔍
code-reviewer
Code Quality
Zero-context code review — correctness, security, performance, readability
🤖
creative-director
Specialist
---
🤖
daniel-polymarket-trader
Specialist
---
🤖
file-analyzer
Specialist
---
🏷️
gmc-guy-olane
GMC OLANE
Google Merchant Center product feed optimization for OLANE — title generation, SEO keywords
🏷️
gmc-guy-sokynd
GMC SOKYND
Google Merchant Center product feed optimization for SOKYND — title generation, SEO keywords
🏷️
gmc-guy-zeya
GMC ZEYA
Google Merchant Center product feed optimization for ZEYA — title generation, SEO keywords
🏷️
gmc-guy
GMC SOERA
Google Merchant Center product feed optimization for SOERA — title generation, SEO keywords
🤖
gmc-validator
Specialist
---
🌐
hoss-researcher
Research
Deep web search, file reads, and codebase exploration without polluting main context
🤖
marketing-growth-hacker
Specialist
---
🤖
marketing-instagram-curator
Specialist
---
🤖
marketing-seo-specialist
Specialist
---
🤖
marketing-tiktok-strategist
Specialist
---
🤖
marketing-twitter-engager
Specialist
---
🤖
paid-media-auditor
Specialist
---
🤖
paid-media-creative-strategist
Specialist
---
🤖
paid-media-paid-social-strategist
Specialist
---
🤖
paid-media-ppc-strategist
Specialist
---
🤖
paid-media-programmatic-buyer
Specialist
---
🤖
paid-media-search-query-analyst
Specialist
---
🤖
paid-media-tracking-specialist
Specialist
---
🤖
parallel-worker
Specialist
---
🧪
qa-agent-code
Quality Assurance
Generates tests for code snippets, runs them, and reports pass/fail results
🤖
sergio-google-ads-agent
Specialist
---
🤖
specialized-document-generator
Specialist
---
🤖
specialized-mcp-builder
Specialist
---
🤖
sunny-polymarket-trader
Specialist
---
🤖
supply-chain-strategist
Specialist
---
🤖
test-runner
Specialist
---
🤖
zuck-meta-ads-agent
Specialist
---

💬 Telegram Commands

/start Show bot status and current AI model
/clear Reset conversation history for a fresh start
/model Show current AI model and switch options
/model claude Switch to Claude Sonnet (Anthropic API, paid, best quality)
/model llama Switch to Llama 3.2 3B (free, runs locally on VPS)
/model list List all Ollama models installed on the VPS
/model mistral:7b Switch to any specific Ollama model by name

🧠 AI Models

Switch between models live in Telegram. Switching always clears conversation history.

Llama 3.2 3B
Free
Command/model llama
Best forCasual chat, simple questions
CostFree
Speed~10 tok/s (no GPU)

🛠 Available Skills

Send a natural language request — Javis picks the right skill automatically.

Lead Generation

scrape-leads
"Scrape 200 leads for plumbers in Texas"
gmaps-leads
"Find 50 HVAC companies in Berlin from Google Maps"
classify-leads
"Classify these leads as agency or SaaS"
casualize-names
"Casualize the company names in my leads sheet"

Email & Campaigns

welcome-email
"Send the welcome sequence to new client John"
gmail-inbox
"Check my inbox and summarize action items"
gmail-label
"Label my Gmail inbox into categories"

Sales & Proposals

create-proposal
"Generate a proposal for client Acme GmbH"
upwork-apply
"Find Upwork jobs for video editing and write proposals"

Content & Video

video-edit
"Edit my video, remove silences"
youtube-outliers
"Find viral YouTube videos in the marketing niche"
cross-niche-outliers
"Find viral videos from adjacent niches"
title-variants
"Generate 10 title variants for my YouTube video"
recreate-thumbnails
"Recreate this thumbnail with my face"
pan-3d-transition
"Add a 3D swivel transition to my video"

Community & Research

skool-monitor
"Check Skool for unread posts"
skool-rag
"Search my Skool community for posts about cold email"
literature-research
"Search PubMed for studies on intermittent fasting"

Client Onboarding & Infrastructure

onboarding-kickoff
"Run full onboarding for new client Acme GmbH"
modal-deploy
"Deploy my Modal webhook"

🚀 GSD Framework

Get Shit Done — a meta-prompting and context engineering system that fights context rot (quality degradation as the context window fills up during long sessions).

What It Does

GSD breaks complex builds into atomic tasks, each executed in a fresh context window. Requirements, roadmaps, and state are persisted in structured markdown files and re-injected per task — so task 10 executes as well as task 1.

Workflow

/gsd:new-project
Initialize — gather requirements, spawn research agents, create roadmap
/gsd:discuss-phase
Discuss — capture implementation preferences before planning
/gsd:plan-phase
Plan — research approaches, create atomic task plans, verify against requirements
/gsd:execute-phase
Execute — run plans in parallel waves with fresh context, commit after each task
/gsd:verify-work
Verify — manual testing with automated diagnosis of failures
/gsd:help
Help — full command reference

When To Use

Use GSDNew features with 10+ steps, multi-file changes, long builds where Claude starts forgetting context
Skip GSDQuick fixes, single-file edits, running existing skills, tasks that fit in one session
Install: npx get-shit-done-cc@latest
Source: github.com/gsd-build/get-shit-done
Version: 1.22.4 (installed globally at ~/.claude)

Superpowers

A composable skills framework for structured development workflows — brainstorming, planning, TDD, code review, and subagent orchestration.

Skills Included

brainstorming
Refine project goals through structured dialogue
writing-plans
Break work into bite-sized 2-5 minute tasks
executing-plans
Execute planned tasks with subagent delegation
test-driven-development
RED-GREEN-REFACTOR cycle enforcement
requesting-code-review
Validate implementation against the plan
systematic-debugging
Structured approach to finding and fixing bugs
subagent-driven-development
Orchestrate parallel subagents for complex tasks
using-git-worktrees
Isolated development environments per feature
finishing-a-development-branch
Handle merging and cleanup
verification-before-completion
Evidence-based checks before marking done
Install: claude plugin marketplace add obra/superpowers-marketplace then claude plugin install superpowers@superpowers-marketplace
Source: github.com/obra/superpowers
Version: 4.3.1 (Claude Code plugin)

📊 Token Usage Dashboard

Live data from bot API calls · Updated every 15 min

Today
0
$0.000
This Week
0
$0.000
This Month
0
$0.000
Total API Calls
18
2.0M tokens

Usage by Source

Chat
1.1M
9 calls
Scheduled
0
0 calls
Heartbeat
869.3K
9 calls

Last 7 Days

0
0
0
0
0
0
0
07-31
08-01
08-02
08-03
08-04
08-05
08-06

Recent API Calls

TimeSourceModelInputOutputEst. Cost
2026-06-01 12:38chatSonnet130,207112$0.3923
2026-06-01 12:38chatSonnet128,231117$0.3864
2026-06-01 12:38chatSonnet127,41482$0.3835
2026-06-01 12:38chatSonnet127,14675$0.3826
2026-06-01 12:38chatSonnet127,04385$0.3824
2026-06-01 11:05chatSonnet127,080408$0.3874
2026-06-01 11:04chatSonnet126,99663$0.3819
2026-03-18 13:06heartbeatSonnet97,876231$0.2971
2026-03-18 13:06heartbeatSonnet97,336125$0.2939
2026-03-18 13:06heartbeatSonnet97,139178$0.2941
2026-03-18 13:06heartbeatSonnet96,498174$0.2921
2026-03-18 13:06heartbeatSonnet96,268103$0.2903
2026-03-18 13:05heartbeatSonnet95,983195$0.2909
2026-03-18 13:05heartbeatSonnet95,760192$0.2902
2026-03-18 13:05heartbeatSonnet95,611119$0.2886
💡 Set a spending limit at console.anthropic.com → Billing → Usage limits to avoid surprises. Llama calls (local) are free and not tracked here.

🛍 Shopify (SOERA)

Live connection to the SOERA Shopify store via MCP. Javis can read and manage products, orders, and customers directly.

Store
yx2v3t-0g.myshopify.com
Auth
OAuth (auto-refresh 24h)

Available Tools

ToolDescriptionScopes
get-productsList/search productsread_products
get-product-by-idGet product details + variantsread_products
create-productCreate new productwrite_products
update-productUpdate product fieldswrite_products
delete-productDelete a productwrite_products
get-ordersList orders by statusread_orders
get-order-by-idGet order detailsread_orders
update-orderUpdate order fieldswrite_orders
get-customersList/search customersread_customers
get-customer-ordersGet orders for a customerread_orders
update-customerUpdate customer infowrite_customers

Current Scopes

read_inventory, read_orders, read_products — add write scopes in the Dev Dashboard under Openclaw app → Configuration → API access scopes.

Example Prompts

Products
"Show me all SOERA products with inventory"
Orders
"List recent orders from the store"
Customers
"Find customer by email"
Inventory
"Which products are running low on stock?"

MCP Config

// .mcp.json
"shopify": {
  "command": "npx",
  "args": ["shopify-mcp"],
  "env": {
    "SHOPIFY_CLIENT_ID": "...",
    "SHOPIFY_CLIENT_SECRET": "...",
    "MYSHOPIFY_DOMAIN": "yx2v3t-0g.myshopify.com"
  }
}

🌐 Web Services

All services run on the VPS behind nginx reverse proxy with SSL (Let's Encrypt, auto-renewing). Domain: adzira-ecom.com (GoDaddy).

ServiceURLPortAuth
Mission Control javis.adzira-ecom.com 8080 Login page → basic auth
Meta Ads Uploader ads.adzira-ecom.com 3000 Login page → basic auth

Stack

ComponentDetails
Reverse Proxynginx 1.24 — routes subdomains to localhost ports
SSLLet's Encrypt via certbot — auto-renews every 90 days
AuthCustom login pages + nginx basic auth on protected paths
Login files/var/www/javis/login.html & ads-login.html
Nginx configs/etc/nginx/sites-available/*.adzira-ecom.com

Adding a new subdomain

# 1. Add A record on GoDaddy: name=SUBDOMAIN, value=187.77.139.236
# 2. Create nginx config
sudo cp /etc/nginx/sites-available/javis.adzira-ecom.com /etc/nginx/sites-available/NEW.adzira-ecom.com
# 3. Edit server_name and proxy_pass port
# 4. Enable and get SSL
sudo ln -s /etc/nginx/sites-available/NEW.adzira-ecom.com /etc/nginx/sites-enabled/
sudo certbot --nginx -d NEW.adzira-ecom.com

🖥 VPS Details

ItemValue
ProviderHostinger
IP Address72.61.207.250
SSH Userroot
Project Path/root/claude-ai/
SSH Key (Mac)~/.ssh/claude_vps
RAM7.8 GB
CPUs2 cores
Disk96 GB

🚀 Deployment & Maintenance

SSH into VPS

ssh -i ~/.ssh/claude_vps root@72.61.207.250

Deploy local changes

./execution/deploy.sh root@72.61.207.250

Auto-sync while working

./execution/watch_and_sync.sh

Leave running in a terminal. Detects file changes every 5s, rsyncs to VPS, restarts bot on .py changes.

Bot management

# Check status
ssh -i ~/.ssh/claude_vps root@72.61.207.250 "systemctl status claude-bot"

# Watch live logs
ssh -i ~/.ssh/claude_vps root@72.61.207.250 "journalctl -fu claude-bot"

# Restart bot
ssh -i ~/.ssh/claude_vps root@72.61.207.250 "systemctl restart claude-bot"

Install a new Ollama model

ssh -i ~/.ssh/claude_vps root@72.61.207.250 "ollama pull mistral:7b"

Then switch in Telegram with /model mistral:7b.

📁 File Structure

CLAUDE AI/
├── JAVIS_USER_GUIDE.md          # Markdown source of this guide
├── .env.example                 # Template for environment variables
├── execution/
│   ├── telegram_bot.py          # Main bot
│   ├── deploy.sh                # Deploy to VPS
│   ├── watch_and_sync.sh        # Auto-sync file watcher
│   ├── requirements_bot.txt     # Python dependencies
│   ├── claude-bot.service       # Systemd service definition
│   └── javis_guide.html         # This website
└── .claude/
    ├── skills/                  # 27 skills (SKILL.md + scripts/)
    ├── agents/                  # Subagent definitions
    └── rules/                   # Operating principles

🔐 Security

⚠️ The SSH key ~/.ssh/claude_vps gives full root access to your VPS. The bot only responds to Telegram user ID 1475855083 (your account).
ItemAction if compromised
SSH key (~/.ssh/claude_vps)Remove from VPS authorized_keys, regenerate
Telegram bot tokenMessage @BotFather → /revoke, update .env on VPS
Anthropic API keyRotate at console.anthropic.com, update .env on VPS
.env fileNever commit to git — already in .gitignore

Update .env on VPS

ssh -i ~/.ssh/claude_vps root@72.61.207.250 "nano /root/claude-ai/.env && systemctl restart claude-bot"

🔧 Troubleshooting

ProblemFix
Bot not respondingsystemctl status claude-bot — check if running
Bot crashedjournalctl -fu claude-bot — read the error
Skill script failsCheck .env has the required API key for that skill
Llama is very slowNormal — no GPU. Use /model claude for heavy tasks
"Unauthorized" in TelegramSomeone else messaged your bot — only your user ID works
Changes not reflectingRun ./execution/deploy.sh root@72.61.207.250