We were up for a technical challenge at Hack Princeton and wanted to fully show the powers and capabilities of AI agents. We also looked at Alpaca, a YC startup that provides stock and crypto API brokerage services. Our goal was to build on this idea and provide users with more than just data — we wanted to offer data, a research agent to help build strategies, and an agent that runs tests in a sandbox environment to validate them. What would normally take hours can now be done in a minute.

We offer three main agentic services:

1. Strategy Builder & Backtester – Describe a strategy or thesis in natural language. The agent understands, builds, and backtests it against historical data and benchmarks across multiple simulations, returning performance metrics and graphs.
2. Research Agent – Chat with an agent that crawls the web, scrapes sites, and performs complex natural language queries on X to gather the latest financial news.
3. Signal Agent – Define signal conditions for monitoring. Once a signal triggers, the agent automatically analyzes it and reports findings in real time.

Together, these services help users develop and test trading strategies faster than ever before.

• Dedalus SDK was used to orchestrate AI agents and workflows.
• X API was integrated for real-time financial news and signal data.
• Agents used yfinance and in-process code execution for backtesting.
• We built MCP servers for web search, scraping, and a custom X MCP server for complex research queries.
• Finally, we designed a webhook-like architecture to push real-time signals from X into Dedalus agents, allowing live analysis.

1. Webhooks: X does not provide webhook endpoints without an enterprise subscription.
2. Invocation Limits: Dedalus agents typically require direct user input to run, not event-based signals.

To overcome this, we engineered a custom webhook-like workaround to handle signal-based agent invocations.

• Building a fault-tolerant backtesting pipeline that orchestrates four services within one agent using smart tool chaining in Dedalus.
• Implementing real-time signal analysis with a creative workaround to bypass API and SDK limitations.

• How to design and implement a custom backtesting engine from scratch.
• How to chain AI tools effectively to perform multi-step, domain-specific tasks.
• That even complex workflows can be automated efficiently through agentic reasoning and orchestration.

• Expanding to include Monte Carlo simulations and complex signal definitions.
• Integrating more data sources for richer backtesting and research.
• Connecting with Alpaca’s brokerage API to enable live trading directly from validated strategies.

• Claude
• Dedalus
• Grok
• MongoDB
• Next.js
• OpenAI
• Python