How do I deploy AI-Trader on Railway?

You can deploy AI-Trader on Railway by clicking the "Deploy Now" button on this page. Railway will automatically set up all the necessary services and configurations for you.

What are the system requirements?

Railway handles all the infrastructure requirements. You only need a web browser to deploy and manage your application.

Is this template free to use?

Yes, this template is free to use within your existing Railway account.

Can I customize this template?

Yes, you can fully customize this template. After deployment, you will have full control of what you've deployed in your Railway project.

How do I get support for this template?

You can get support through our community forum: ai-trader-2b0f423c, or click "Get template help" for this template for more assistance.

Deploy and Host AI-Trader on Railway

AI-Trader is an autonomous trading benchmark that lets multiple LLM agents compete in historical market simulations. Agents research market context, analyze price data, and execute simulated trades using a tool-driven (MCP) architecture. It includes a lightweight web dashboard to visualize portfolio performance, trades, and leaderboard comparisons across models.

About Hosting AI-Trader

This template deploys AI-Trader as a single Railway service using Docker. The container starts all MCP tool services (math, search, trade execution, local price lookup, and crypto tools), runs a configured trading simulation, and serves the static UI dashboard over Railway’s public HTTP port.

A persistent Railway Volume is mounted at /app/data to store datasets and run outputs (price caches, agent positions, transactions, and logs), ensuring results persist across restarts and redeploys. To run real agent decisions, you must provide an OpenAI-compatible API key, and optionally keys for external search and market data providers.

Common Use Cases

Benchmark multiple LLMs on identical historical market data and trading rules
Run reproducible paper-trading simulations over fixed historical periods
Visualize portfolio evolution, trade actions, and model performance in a web dashboard

Dependencies for AI-Trader Hosting

An OpenAI-compatible API key (required for the trading agent’s LLM calls)
A persistent Railway Volume mounted at /app/data (required to store datasets and results)

Deployment Dependencies

AI-Trader upstream repository: https://github.com/HKUDS/AI-Trader
OpenAI API keys: https://platform.openai.com/api-keys
Jina AI (optional, for market/news search): https://jina.ai
Alpha Vantage (optional, for stock price data refresh): https://www.alphavantage.co
Railway Templates reference: https://docs.railway.com/reference/templates
Deploy a Template guide: https://docs.railway.com/guides/deploy

Implementation Details

Environment variables used by this template:

END_DATE="2026-01-12" # Simulation end date (YYYY-MM-DD)
INIT_DATE="2026-01-01" # Simulation start date (YYYY-MM-DD)
CONFIG_PATH="/app/configs/default_config.json" # Scenario config to run (US market, daily). Use /app/configs/default_hour_config.json for hourly runs
JINA_API_KEY="YOUR_JINA_KEY" # Jina API key used by the Search/market-intelligence tool (leave empty to disable external search)
AGENT_MAX_STEP="30" # Max number of reasoning/action steps the agent can take per trading day (higher = more tool calls/cost)
MATH_HTTP_PORT="8000" # MCP Math service port (calculations/metrics)
OPENAI_API_KEY="YOUR_OPENAI_KEY" # OpenAI API key (required to run the agent)
OPENAI_API_BASE="https://api.openai.com/v1" # OpenAI-compatible API base URL (must end with /v1)
TRADE_HTTP_PORT="8002" # MCP TradeTools service port (buy/sell simulation)
CRYPTO_HTTP_PORT="8005" # MCP CryptoTradeTools service port (crypto tools)
RESET_AGENT_DATA="0" # Set to 1 to wipe ./data/agent_data before running a new simulation
RUNTIME_ENV_PATH="/app/runtime_env.json" # Runtime environment config file path used by the app (generated/consumed at startup)
SEARCH_HTTP_PORT="8001" # MCP Search/market intelligence service port (news/search)
FORCE_REFRESH_DATA="0" # Set to 1 to force re-download/refresh of price data on boot (can hit API rate limits)
GETPRICE_HTTP_PORT="8003" # MCP LocalPrices service port (local price lookup)
ALPHAVANTAGE_API_KEY="YOUR_ALPHA_VANTAGE_KEY" # Alpha Vantage API key (used to download/update stock prices when local cache is missing)

Notes:

The UI is served on Railway’s assigned $PORT (public), while MCP tool ports are internal-only.
Mounting a persistent volume at /app/data is required to retain datasets and agent outputs.
Free Alpha Vantage keys are rate-limited; large universes (e.g., Nasdaq-100) may require a paid plan or a reduced symbol set.

Why Deploy AI-Trader on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying AI-Trader on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.