Alpaca Markets

Commission-free brokerage API for programmatic stock trading. Used by Nate Herk to build a fully autonomous Claude Code cloud-routine trading agent named "Bull".

Key facts

• Type: Trading brokerage API
• Status: Active
• Paper trading: $100K virtual account for testing without real funds
• Live trading: real-money account via same API
• MCP server available: yes — can be plugged directly into Claude Code
• Key endpoints: market data, order placement (market, limit, stop), portfolio management, account status

What it does in this ecosystem

Alpaca + Perplexity API + Claude Code cloud routines = a five-routine autonomous trading system:

1. Pre-market: Perplexity API gathers news → identify 3 stocks, generate trading plan → write to today.md

2. Market open: read today.md, place trades via Alpaca, write executed orders to memory files

3. Midday: check positions, adjust stop-losses, write risk management decisions

4. End-of-day: journal performance, write lessons to journal.md

5. Friday: weekly review and strategy update

All five routines share memory via committed files in the same GitHub repo. ^[src-010]

^[src-086] adds a real-money benchmark variant: Nate's OpenClaw trading bot "Bull" used Alpaca to manage a $10,000 account for 30 days, make buy/sell/stop decisions, and compare results against both Salmon's separate agent and the S&P 500 baseline. The experiment is useful as an Autonomous Trading Agents case study, not as evidence that unattended AI trading is safe or reliably profitable.

Related

• See also: Claude Code, Perplexity API, OpenClaw
• Concepts: Claude Code Cloud Routines, Stateless Agent Memory Pattern, Autonomous Trading Agents, Agent Security Boundaries

Source references

• ^[src-010] Nate Herk — Cloud agents & model releases cluster (2026-04-14 to 2026-04-17)
• ^[src-086] Nate Herk — "I Gave OpenClaw $10,000 to Trade Stocks" (2026-04-09)