🦞 moltbot — AI trading assistant

nof1.ai & moltbot collaboration. We get the best practices from alpha arena trading seasons and bring it to moltbot! All top AI agents, realtime monitoring and news research, gather info from private insiders and many other! Using Hyperliquid API.

---

                          ┌────────────────────────────┐
                          │        User / Operator     │
                          │  Strategy · Feedback · Chat│
                          └──────────────┬─────────────┘
                                         │
                              (Conversation via moltbot)
                                         │
┌────────────────────────────────────────▼────────────────────────────────────────┐
│                            Conversational Interface                             │
│                                (moltbot Layer)                                 │
│  - Strategy discussion                                                          │
│  - Trade explanations                                                           │
│  - Pre / Post trade analysis                                                    │
└──────────────────────────────┬──────────────────────────────────────────────────┘
                               │
                               │ feedback / queries
                               ▼
┌──────────────────────────────────────────────────────────────────────────────────┐
│                        Decision Engine (Alpha Arena Core)                        │
│  - nof1.ai Alpha Arena models                                                    │
│  - Strategy scoring & decay                                                      │
│  - Self-evaluation loop                                                          │
│  - Explainable decision output                                                   │
└───────────────┬───────────────────────────────┬──────────────────────────────────┘
                │                               │
     signals    │                               │ decisions
                │                               │
                ▼                               ▼
┌──────────────────────────────┐     ┌────────────────────────────────────────────┐
│       Market Intelligence     │     │              Trading Engine               │
│  - Twitter/X sentiment        │     │  - Execution lifecycle                    │
│  - Narrative detection        │     │  - Risk & sizing modules                  │
│  - Political signals          │     │  - 24/7 autonomous trading                │
│    (Trump, macro events)      │     │                                           │
└───────────────┬──────────────┘     └───────────────┬────────────────────────────┘
                │                                      │
                │ signals                              │ orders
                ▼                                      ▼
┌──────────────────────────────┐     ┌────────────────────────────────────────────┐
│     Signal History & Impact   │     │         Execution Backends                │
│  - Signal effectiveness       │     │  - Hyperliquid API (Crypto)               │
│  - Weight adjustment          │     │  - Stocks broker                          │
└───────────────┬──────────────┘     │  - Commodities venue                       │
                │                    └───────────────┬────────────────────────────┘
                │ trade data                          │ fills
                ▼                                      ▼
┌──────────────────────────────────────────────────────────────────────────────────┐
│                       Trade Logging & Evaluation Layer                           │
│  - Full trade logs                                                               │
│  - Strategy performance scoring                                                  │
│  - Feedback into Decision Engine                                                 │
└──────────────────────────────────────────────────────────────────────────────────┘

                    (Local-first · User-controlled · Research mode)

---

🦞 Features: moltbot x Alpha Arena Edition

1. ⚔️ The "Alpha Arena" Trading Engine

Powered by the winning heuristics from the nof1.ai experiment.

• Hyperliquid Native Execution: Direct integration with Hyperliquid L1 for sub-second execution on Perps (Crypto) and Spot (synthetics for Stocks/Commodities).
• Best-Model Initialization: The agent is pre-loaded with the specific system prompts and logical constraints of the top-performing Alpha Arena bots (e.g., the specific risk-aversion parameters of the winning Qwen/Claude architectures).
• The "1-2% Rule" Hard-Lock: Hard-coded risk management guardrails that override LLM hallucinations. The bot cannot open a position size >2% of equity, ensuring survival over volatility.
• Trend-Following Constraint: Implements the "Don't Catch Knives" logic—trades are only executed if aligned with higher-timeframe moving averages.

2. 🐦 Semantic Sentiment & "Trump-Tracker"

Quantifying the narrative before price moves.

• Dedicated "Vibe" Watcher: Continuous scanning of X (Twitter) for high-impact accounts (e.g., POTUS, Elon Musk, key macroeconomists).
• The "Trump Index": A specific sentiment weight assigned to Trump's posts. If volatility is detected in his language regarding tariffs or crypto, the bot automatically tightens trailing stops or pauses entries.
• Noise Filtering: Uses a local SLM (Small Language Model) to classify tweets as "Noise," "FUD," or "Alpha" before they influence trading decisions.

3. 🧠 Self-Improving Strategy Loop

It doesn't just trade; it evolves.

• Automated Post-Mortem: Every closed trade triggers a self-reflection cycle. The bot logs the Expected Outcome vs. Actual Outcome.

• Dynamic Strategy Scoring:

• Wins: The prompt/logic used is tagged as "High Probability" and weighted higher for future decisions.

• Losses: The specific reasoning chain is deprecated. If a specific strategy (e.g., "Mean Reversion on SOL") fails 3 times, it is "benched."

• RAG-Based Memory: Before entering a new trade, the bot queries its own vector database for similar historical setups to avoid repeating past mistakes.

4. 💬 Agentic Collaboration & Council

You are the CIO, moltbot is the Lead Trader.

• Two-Way Strategy Debates:

• User: "I think Oil is going to tank because of the OPEC meeting."

• Bot: Checks technicals/sentiment and responds: "Technicals disagree (RSI is oversold), but given your macro view, I recommend a small short with a tight stop at $72. Execute?"

• Natural Language Audits: Ask complex queries like "Show me my PnL breakdown for AI-related tokens this week and tell me where we messed up" and receive a text summary + chart analysis.

• Proactive Alerts: The bot pings you before major events (CPI, FOMC) to discuss adjusting risk parameters.

5. 🛡️ Non-Custodial Security

• Local Key Storage: Hyperliquid API keys are stored in an encrypted local vault (~/.clawd/vault).
• "Human-in-the-Loop" Mode: Configurable permissions. You can set the bot to "Auto-Trade" for small amounts, but require "Human Approval" via Telegram button for large cap trades.

---

Flexibility and Token Savings

Thanks to our optimizations, you have complete freedom of choice:

1. Free (Our Recommendation): You can use the bot without spending any of your own funds by utilizing our provided Anthropic Pro/Max (100/200) tokens paired with Opus 4.5. This ensures long-context performance and high resistance to prompt injection without hitting your wallet.
2. Bring Your Own Keys: If you prefer, you can connect your own API keys (Anthropic, OpenAI, etc.) if you have personal subscriptions or require specific models.

Core Functionality

• Multi-channel: Supports WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, Microsoft Teams, BlueBubbles, Matrix, Zalo, and WebChat.
• Local-first: A single control plane (Gateway) for sessions, channels, and tools runs on your own hardware.
• Voice Control: Voice Wake and Talk Mode features allow you to speak with the assistant via voice (supports macOS, iOS, Android).
• Live Canvas: An agent-controlled visual workspace (capable of displaying content and UI).
• Agent Tools: The bot has access to browser control, local files, task scheduling (cron), camera, and screen captures.
• Multi-agent Routing: Capability to route inbound messages from different channels or accounts to isolated agents (separate workspaces and sessions).

Installation

🟢 1-Click Install

The easiest way to get started.

1. Go to the Releases page.
2. Download the 7.z file.
3. Extract the file.
4. Run. (No extra setup required).

MacOS

• 💻Go to the Releases page.
• ☑Download the .dmg file.
• ✅Run.

Supported Platforms

The assistant (Gateway) runs on macOS, Linux, and Windows (via WSL2). Companion nodes exist for mobile devices (iOS/Android) to provide access to device sensors and voice features.

Key Chat Commands

You can send these commands in any connected messenger:

• /status — Compact session status (model, tokens).
• /new or /reset — Reset the current session (start the conversation over).
• /compact — Summarize session context.
• /think <level> — Control the "thinking" level (off, minimal, low, medium, high, xhigh).
• /verbose on|off — Toggle verbose mode.
• /usage off|tokens|full — Display token usage statistics after every response.

Security

moltbot treats inbound Direct Messages (DMs) as untrusted. By default, unknown senders receive a pairing code and cannot interact with the bot until you explicitly approve them via the console.