By sw.lee — Feb 11, 2026

TBOT Controller — TradingBoat as an OpenClaw Skill

Introduction

OpenClaw is powerful. But most people use it for lightweight automation.

What happens when you connect OpenClaw to a real trading runtime — with clear boundaries, deterministic behavior, and explicit control?

TBOT Controller is an OpenClaw skill that connects natural language directly to a production-ready TradingBoat runtime.

It allows you to execute trading actions safely using plain English.

This is not “LLM trading directly.”

This is controlled automation.

How It Works

You speak naturally:

close 50 NFLX

The skill converts that instruction into a structured, validated trading signal and sends it to the TradingBoat runtime.

The runtime then handles broker execution.

The AI interprets.

The runtime executes.

Each layer has a clear responsibility.

Why This Matters

Most AI trading experiments fail because they blur boundaries.

TBOT Controller keeps them clean:

No direct broker access from the LLM
No hidden credentials inside prompts
No uncontrolled execution
Deterministic webhook-based signals

Architecture Overview

The system is intentionally layered.

User (natural language)
↓
OpenClaw
↓
tbot-controller skill
↓
openclaw-on-tradingboat runtime
↓
IBKR Broker

Installation

Install OpenClaw

If you don’t already have OpenClaw installed, install it globally via npm:

npm install -g openclaw

Install the skill:

npx clawhub@latest install tbot-controller

Clone and run the TradingBoat runtime:

git clone https://github.com/PlusGenie/openclaw-on-tradingboat.git
cd openclaw-on-tradingboat
docker compose up -d

(Optional but recommended)

export TBOT_COMPOSE_DIR="$HOME/develop/github/openclaw-on-tradingboat"

Example Trading Signal

When you say:

close 50 NFLX


{
  "timestamp": 1710000000000,
  "ticker": "NFLX",
  "currency": "USD",
  "timeframe": "1D",
  "clientId": 1,
  "key": "WebhookReceived:123456",
  "contract": "stock",
  "orderRef": "Close_NFLX_50_1710000000000",
  "direction": "strategy.close",
  "metrics": [
    {"name": "qty", "value": 50}
  ]
}

This payload is schema-validated before being sent to the runtime.

Security Model

Clear separation between AI and broker
Webhook-based execution
Secrets stored in runtime .env
Deterministic containerized runtime
No hidden side effects

This is controlled automation — not blind AI execution.

Conclusion

TBOT Controller turns OpenClaw into a structured trading operator.

Natural language becomes executable intent.

And execution remains safe, bounded, and transparent.

TBOT Controller transforms OpenClaw from a general-purpose agent into a controlled trading operator.