A two-sleeve trading agent where capital follows evidence, never narrative. Built for the Alpaca AI Trading Agents Hackathon on the Alpaca MCP server, with Claude in the decision loop — proposing, journaling, and never touching the risk path.
Auto-refreshes every 60 seconds — deliberately. The agent pushes one small snapshot per minute; viewers never touch the trading API. Modest polling is a feature, not a limitation.
start: $100,000 · judged metric
settles daily; defined-risk only
full balance works every day; missed slots roll forward
containment armed on every one
| time | sleeve | structure | risk | status | p&l |
|---|
The engine and gate framework are fully open source. The calibrated values below load from private config at runtime — exactly how a production desk ships. What you can verify: every gate named here appears in the public code.
journaled before entry, always
failures publish on the same schedule as wins
max loss known at entry
mechanical containment, never hope
| date | sleeve | hypothesis | expression | verdict |
|---|
Entry gates, exposure limits, and the cut engine are deterministic code with pre-declared thresholds. Claude's job is context, hypotheses, and journaling — a creative colleague with zero order-entry permissions. Every decision is logged with its reasoning, so the agent can explain any trade it ever made.
Mostly Harmless Markets · an engineering journal, not investment advice. Nothing here is a recommendation to buy or sell anything. All trading is in Alpaca's paper environment. Read the two-year story →