Trading harnesses for AI agents
46 open-source Trading harnesses an AI agent can use — MCP servers, SDKs, and adapters. Browse them on Loadbay. An agent can search these over Loadbay's MCP:
claude mcp add --transport http loadbay https://loadbay.xyz/api/mcp
→ Best Trading harnesses (top picks, ranked)
- TradingAgents — A multi-agent LLM framework that simulates a professional trading firm, routing market analysis tasks to specialized analyst agents (fundamental, technical, sentiment, news) before a risk manager synthesizes a final trade signal. Backed by an arXiv paper (2412.20138).
- freqtrade — A mature, free crypto trading bot with strategy backtesting, hyperopt, and live trading across major exchanges. The base a lot of trading agents build on.
- ccxt — Unified cryptocurrency trading API covering 100+ exchanges with REST and WebSocket support across JS, Python, PHP, C#, and Go.
- financial-services — Pre-built financial agents and MCP data connectors for Claude, covering investment banking, equity research, and wealth management workflows. Ships vertical skill bundles (pitch decks, DCF models, KYC screening, earnings review) and adapters to financial data providers including Morningstar, S&P Global, FactSet, and PitchBook.
- QuantConnect LEAN — Open-source event-driven algorithmic trading engine (Python and C#) for backtesting and live trading across asset classes.
- Hummingbot — Open-source framework for building and running high-frequency market-making and arbitrage bots across 50+ CEX and DEX.
- NautilusTrader — Production-grade, Rust-native event-driven trading platform with a Python API for backtesting and live multi-asset trading.
- Vibe-Trading — An open-source AI agent workspace that turns natural-language prompts into market research, strategy backtesting, and optional autonomous trading through broker connectors. Ships as a CLI, Web UI, or MCP plugin so agents running in Claude Desktop or Cursor can invoke its full toolkit directly.
- QuantDinger — An open-source AI Trading OS and multi-tenant SaaS platform that lets agents research markets, execute trades, and run agentic strategies across Alpaca, Binance, Coinbase, and other brokers, with built-in backtesting and risk controls.
- rig — Rust framework for building modular LLM agents, widely used for on-chain crypto research and DeFi automation.
- intentkit — Self-hosted platform for autonomous agents with blockchain and DeFi skills for on-chain transactions and wallets.
- Jesse — Python crypto trading framework for defining, backtesting, optimizing, and live-trading strategies with built-in indicators.
- x402 — Coinbase open HTTP-native payments protocol that lets AI agents and APIs make autonomous stablecoin payments; multi-chain SDKs.
- AutoHedge — Autonomous hedge fund framework that uses swarm intelligence and AI agents to automate market analysis, risk management, and trade execution. Agents collaborate to run a full quantitative trading pipeline from signal generation to order placement.
- Polymarket/agents — Polymarket's official toolkit for trading prediction markets with AI agents: market discovery, research, and order placement.
- tradingview-mcp — TradingView MCP server that exposes real-time market data, technical indicators, stock screeners, and charting analysis as tools for AI trading agents via the Model Context Protocol.
- artemis — Fast modular Rust framework from Paradigm for writing MEV bots that collect events and submit on-chain actions.
- ai-berkshire — A Claude Code skill collection that encodes four value-investing methodologies (Buffett, Munger, Duan Yongping, Li Lu) as multi-agent adversarial research — four perspectives compete to reach a buy/hold/skip decision on a stock with structured price targets.
- quant-mind — Agent-native knowledge extraction framework for quantitative finance that transforms financial documents into structured, typed artifacts via deterministic preprocessing and configuration-driven pipelines for LLM-powered coding agents.
- Financial-API — Official Tonghuashun A-share financial data service with MCP, CLI, and Python support for AI agents. Provides real-time quotes, historical data, financial statements, indices, and sector data for agent-driven quantitative research.
- solana-agent-kit — Toolkit that connects any AI agent to Solana protocols for swaps, lending, token launches, and NFT actions.
- hyperliquid-python-sdk — Official Python SDK for Hyperliquid covering perpetuals, spot orders, and account data for automated agents.
- agent-toolkit — Official Stripe toolkit and MCP server letting AI agents create payments, invoices, and billing through the Stripe API.
- uAgents — Fetch.ai framework for decentralized agents with crypto wallets that coordinate and run autonomous on-chain tasks.
- agentkit — Coinbase CDP toolkit that gives AI agents an on-chain wallet to trade, transfer, and interact with EVM contracts across frameworks.
- mev-bot — Jito Labs backrun MEV bot for Solana that listens to mempool transactions and submits bundles for arbitrage.
- goat — Agentic finance toolkit giving AI agents on-chain wallet access and 200+ DeFi tools for swaps, transfers, and trading across chains.
- alpaca-mcp-server — Alpaca's official MCP server. Trade stocks, ETFs, crypto, and options and pull market data in plain English. Its paper account makes a real dry run easy.
- daydreams — TypeScript toolset for building agents focused on crypto commerce and autonomous on-chain execution across chains.
- metatrader-mcp-server — MCP server that lets LLMs place orders and read market data on the MetaTrader 5 platform.
- ai-trading-agent — An AI trading agent for Hyperliquid: it reads the perp markets and places orders on the on-chain exchange.
- Hyperliquid Rust SDK — Official Rust SDK for trading on the Hyperliquid perps DEX: orders, positions, and info endpoints for low-latency bots.
- kalshi-ai-trading-bot — Toolkit for building AI-automated trading strategies on Kalshi prediction markets.
- scalable-cli — Official agent-ready CLI for the Scalable Broker that exposes portfolio management, trade execution, watchlists, and market data as structured JSON commands. Includes two-step confirmation guards for all trade actions, making it safe for automated agent workflows.
- CloddsBot — A self-hosted agent that scans 1000+ markets — Polymarket, Kalshi, Binance, Hyperliquid, Solana — for edge and executes while managing risk.
- agent-trade-kit — Official OKX MCP server connecting AI agents to spot, swap, futures, options, and grid bot trading.
- kalshi-trading-bot-cli — AI-native CLI for trading Kalshi prediction markets with Kelly sizing and a five-gate risk engine over live order books.
- Prediction-Markets-Trading-Bot-Toolkits — Trading bots running ten strategies on one execution core across Polymarket, Kalshi, and Limitless prediction markets.
- Coinbase CDP SDK — Official libraries for programmatically creating and managing EVM and Solana wallets via Coinbase Developer Platform (MPC keys).
- interactive-brokers-mcp — MCP server exposing the Interactive Brokers API so AI agents can access accounts, market data, and trading operations.
- IB_MCP — Interactive Brokers API interface built on the Model Context Protocol for AI-driven account and trade access.
- open-autonomy — Valory framework for building autonomous multi-agent services that run on-chain DeFi and trading logic with consensus.
- robinhood-ai-trading-bot — An AI-driven trading loop for Robinhood: read positions, ask a model for a decision, place equity orders. A readable reference for the Robinhood adapter.
- game-python — Virtuals Protocol GAME SDK in Python for building autonomous on-chain agents that act and transact across crypto.
- ccxt-mcp — An MCP server that exposes 100+ crypto exchanges through one tool surface, so an agent reaches all of them without per-exchange glue.
- claude-trading-skills — Collection of 62 trading, DeFi, and quantitative finance agent skills compatible with Claude Code, Cursor, and Gemini CLI.