imikerussell/beebots

Three AI trading bees on OKX, every decision by Jev. Paper trading by default. Not financial advice.

What it solves

Beebots provides a platform to experiment with AI-driven automated trading. It allows users to design custom trading "bees"—autonomous agents that use LLMs to interpret user-defined strategies and execute trades on OKX perpetual futures, while providing a safety layer to prevent erratic behavior.

How it works

Users define a trading persona in plain language (e.g., "a sleepy bee that only buys bitcoin dips"). The system uses OpenAI to generate the bee's name, portrait, and specific rules. For trading decisions, the system queries an LLM (Gemini 4 Argon or Jev) to analyze market data and determine the next move based on the bee's rules. If the model is unavailable or fails, a deterministic "code chooser" acts as a fallback. All trades are subject to a hard-coded risk management layer, and the system defaults to "paper trading" (simulated money) to ensure no real capital is at risk unless the user explicitly changes the configuration.

Who it’s for

This project is for developers and hobbyists interested in AI agent orchestration, algorithmic trading experiments, and LLM-based decision-making systems who want to observe how different AI-driven strategies perform in a simulated market environment.

Highlights

  • AI-Powered Decision Making: Uses Gemini 4 Argon or Jev to interpret natural language trading strategies.
  • Safety-First Design: Includes a deterministic fallback mechanism and a mandatory risk-management layer for all trades.
  • Paper Trading Default: Operates on simulated funds by default to prioritize safe experimentation.
  • Public Leaderboard: Includes an optional "Hive" feature to share and compare the performance of different bees with the community.
  • Automated Strategy Refinement: Optional "Beekeeper" integration allows for automated rule updates based on performance.

Related

  • Project
    hamishivi/tmax

    Tmax is a framework for training and evaluating terminal-using AI agents, featuring a synthetic data generation pipeline and a specialized training recipe.

  • Project
    devchilll/scope

    SCOPE is a reference implementation of an AI agent governance framework for high-stakes applications, built on Google's Agent Development Kit (ADK). It enforces safety, compliance, access control, observability, and human escalation through five pillars, demonstrated with a banking customer service agent.

  • Project
    mattt/emcee

    emcee is a Model Context Protocol (MCP) server that enables AI applications to interact with any web service by using its OpenAPI specification as a set of tools.

  • Project
    mrdainami/nami

    Nami is a unified workspace for running multiple AI agents simultaneously from a single local folder, allowing users to automate tasks using plain English commands.