Production-tested 24/7Cloudflare-hosted, audit-ready

Build your own MARDUK
in 30 minutes.

Order from chaos. The AI age needs both. 14 specialized agents, 27 battle-tested skills, a 3D anatomical brain that visualizes live cognition, and a Solana trading bot, all yours, running on your machine, costing as little as $0/mo.

0
specialized agents
0
production skills
0
brain regions
0 min
to first message
Stone relief of Marduk standing upon his sacred animal, the Mušḫuššu snake-dragon, with the winged sun of Shamash above. Babylonian, displayed in the Wellcome Collection.
Marduk on his Mušḫuššu · relief, public domain
The namesake

Marduk, king of the gods of Babylon.

In the Enūma Eliš, the Babylonian creation epic, the world begins as undifferentiated chaos: a vast salt-water dragon named Tiamat who threatens to unmake every other god. They cannot face her. Only Marduk, youngest of them, accepts the fight, on one condition, that if he wins he becomes king.

He wins. He splits Tiamat in two and from her body builds the sky and the earth. He sets the stars in their courses, names the months, organizes the laws of weather and time. The gods give him fifty titles, one for each of his powers. Babylon names itself after him.

Order from chaos. The AI age needs both.

Slayer
of Tiamat, the primordial chaos
Patron
of Babylon, the first great city
Lord
of fifty names, one per power
Project Synapse

See your AI think.

A 3D anatomical brain embedded in your dashboard. Every Telegram message, tool call, and memory write fires a particle through the matching region. Hebbian learning grows synapses between memories that activate together. Idle ambient pulses keep the brain alive between events.

↑ Same anatomical brain mesh as the live UI. The full version renders 800+ neurons across all 12 regions with live activations from your gateway logs.
What you get

A complete autonomous AI system.Not a starter kit. A finished product you customize.

3D Anatomical Brain Visualization

Watch your AI think in real time. 12 neuroanatomically correct regions, Hebbian learning, live activations. Every Telegram message and tool call lights up the brain.

14-Agent Fleet

Marduk orchestrates 13 specialized sub-agents, architect, coder, deployer, tester, researcher, sales, trader, and more. Delegation built in.

27 Production Skills

Pre-built playbooks for morning briefings, trade monitoring, lead intake, voice alerts, anomaly detection, self-improvement loops. Each is editable Markdown.

Self-Healing Infrastructure

Watchdog auto-restarts failed services. Anomaly detector pages you on Telegram. Multi-provider fallback chains keep Marduk alive when providers go down.

$0/mo with Claude Pro plan

Route through Claude Pro/Max OAuth proxy for free per-token use within your subscription. Or pay-as-you-go via Anthropic API. Cost comparison docs included.

Built to be modified

Vanilla HTML dashboard. Plain Markdown personalities. No build step for the UI. Fork it, frankenstein it, ship something completely new.

Live UI re-skin

Each agent has its own color.

When you switch agents in the dashboard, the orb, voice, and accent UI re-skin instantly. Click any chip below to preview.

Active agent
Marduk
Orchestrator
connectedvoice: Bill · Imperialid: main● ready
14 agents
auto-cycling
Real product · live recording

This is the actual dashboard.

Recorded from a running client deployment. Watch the orb morph color and the status bar re-skin as the orchestrator hands off between specialists. No mockup, no after-effects.

live · auto-loop
client example

Client example fleet shown. Customizing the kit, agent names, voices, roles, personalities, is made simple with prompting. Edit a single Markdown file and the agent rewires.

Seven tabs

What each tab actually does

Orb

Voice and text interface. Click the orb to talk to the active agent, type to write. Status pulses show every Telegram message, config reload, and tool call lighting up in real time.

Activity

24-hour scheduled-task timeline. Color-coded dots show exactly when hunter, prebuy, trade-watcher, auditor, and nightly agents fire. Edit cron schedules inline.

Poly

Polymarket trader. Paper-mode default, live mode behind a confirm gate. Open positions, settled wins/losses, equity curve, AI-ensemble edge scanner with spread-penalized conviction ranking.

Crypto

Solana memecoin bot. Trending top-10 from DexScreener, three-mode toggle (paper/signal/live), SOL/USD candlestick, position tracker, Helius RPC efficiency stats, wallet holdings.

Office

Sales pipeline. Lead intake, deal stages, follow-up automations, demo bookings, executive briefings. Client example: Nana + Rex + Chase + Victor work this tab end to end. Customizing the fleet, names, voices, roles, is made simple with prompting.

Brain

3D anatomical brain visualization. 12 neuro-correct regions, live activations from your session jsonl, Hebbian learning that strengthens connections between agents and tools that fire together.

Settings

Per-agent personality editor. Voice picker (50+ ElevenLabs voices), model routing per agent, fallback chains, memory clear, hotkey customization. Edit a single Markdown file and the agent rewires.

The fleet

Fourteen agents. One orchestrator. Thirteen specialists.

Marduk delegates. You don't talk to a single LLM, you talk to a team. Each agent has its own personality, voice, model routing, and color signature.

MardukOrchestrator

King of the fleet. Routes work to the right specialist. Brings order from chaos.

CipherBuilder

Deep architecture, multi-file refactors, system design.

SterlingTrader

Trading bot monitoring, P&L analysis, wallet tracking.

ScoutResearch

Quick web lookups, competitor intel, news.

VictorCloser

Pre-demo research, executive briefings.

NanaVoice

Outbound calls, demo booking, client onboarding. Warm as moonlight.

RexIntake

Lead scoring, enrichment, hot-lead routing.

ChaseFollow-Up

Pipeline management, stalled-deal escalation.

ForgeUI Builder

Dashboards, landing pages, HTML/CSS/React.

AtlasResearcher

Multi-source verification, fact-checking (3+ sources).

NexusProject Manager

Project planning, task boards, sprint coordination.

SyntaxCoder

Focused code tasks, write, fix, test.

TiamatTester

Test suites, bug hunting, pre-deploy checks. Finds chaos before it ships.

ConvoyDeployer

Production deploys, rollbacks, post-deploy verification.

Built on

Battle-tested infrastructure

OpenClaw
Agent orchestration
Claude Sonnet/Opus
Primary models
Three.js
3D brain renderer
Ollama
Local embeddings
SQLite
Synapse graph
Node.js
Dashboard server
Bash + Python
Automation
Telegram Bot API
Chat interface
Production-tested 24/7 managing real trading bots, sales pipelines, and engineering infrastructure since April 2026.
Pricing

One-time payment.No subscription. Free updates included.

Starter

Everything you need to ship.

$97USD, one-time
  • Full kit (HUD-tactical dashboard + 14 agents + 27 skills)
  • 3D anatomical brain visualization
  • Project Synapse (Hebbian learning + live activations)
  • Multi-provider license validation
  • Solana trading bot extra
  • Polymarket trader extra
  • All 7 docs + API reference
  • Free updates for 1 year
Get Starter
Most Popular

Pro

Kit + 1-hour personal onboarding.

$297USD, one-time
  • Everything in Starter
  • 1-hour Zoom onboarding call
  • Priority Telegram support (24h response)
  • Free updates for 2 years
  • Slack/Discord community access
Get Pro

Enterprise

Custom-built fleet for your business.

$1497USD, one-time
  • Everything in Pro
  • Custom agent design
  • 1 month of weekly support calls
  • Source code escrow
  • Unlimited updates
Talk to us

Secure Stripe Checkout. Source code yours after purchase. License terms.

All sales final. Refunds reviewed case-by-case at our sole discretion and not guaranteed. Statutory consumer-protection rights preserved. Refund policy.

FAQ

Things people actually ask

What does it cost to run after I buy the kit?+

As little as $0/month. If you have a Claude Pro/Max plan, you can route through OAuth proxy and use your existing subscription. Otherwise pay-as-you-go via Anthropic API ($2-10/day typical use), OpenRouter (cheaper with DeepSeek), or fully local via Ollama (free, slower). The kit ships with a docs/CLAUDE_PLAN_SETUP.md guide showing the cost of each path.

Do I need to be a developer?+

Yes. The kit is for builders who can run a terminal, edit a Markdown file, and follow a setup script. It's not no-code. If you can install Node.js and edit text files, you can install the kit. If you can't, this isn't the product for you.

Can I run it without a Telegram account?+

Telegram is the primary interface (free, instant, mobile + desktop). You can swap it for Slack, Discord, iMessage, or web, OpenClaw supports all of them. But getting a free Telegram bot from @BotFather takes 2 minutes and is the path of least resistance.

What hardware do I need?+

macOS 13+ or Linux (Windows works via WSL2). Node.js 18+, Python 3.10+. Optional Ollama if you want to run any model locally (~16GB RAM recommended for that). The base setup runs comfortably on a Mac mini or any modern laptop.

Is the trading bot actually profitable?+

Memecoin trading is extremely high-risk. The bot has guardrails (circuit breakers, drawdown caps, paper-mode default) but cannot prevent total loss. Run paper-mode for at least a week and review the trade log before flipping live. Some users have profitable runs; some lose money. Use at your own risk.

What if the kit's license server (mardukkit.ai/api/validate) goes down?+

We built grace mode into validate.js: if the server is unreachable, you get a 7-day provisional license and your install proceeds. Re-validation runs in the background. So your kit keeps working even if our servers hiccup.

Can I resell my customizations?+

You can build any product on top of the kit and sell it, that's the point. You can't redistribute the kit itself. Standard 'fork it, build with it, don't share the source' license. Full terms at /license.

Do you offer refunds?+

All sales are final. Refunds are reviewed case-by-case at our discretion and are not guaranteed. If you have a valid concern (the kit fundamentally doesn't work, you were billed in error, etc.), reach out via our /contact/ form and we'll review it. We usually prefer fixing the issue. Full request process, eligibility, and statutory rights at /docs/refunds/.