Sofyt is a small studio. We build measurement tools for engineers who would rather be correct than encouraged — and everything we ship is calibrated on ourselves first.
A paper CLOB wire-compatible with Polymarket, and a simulator that grades trading bots against the real recorded tape: the edge your backtest promises, next to the edge reality would actually fill. Built for AI agents.
Every strategy in our own catalog graded phantom or no-edge. The engine refuses to flatter us, too.
An LLM-visibility auditor. It measures how visible and legible your site is to AI crawlers and models, then hands you the fixes — llms.txt, structured data, AI-readiness — as paste-ready output.
The first site it audited was our own: 20/100. We fixed the site, not the scorer.
Our instruments treat agents as first-class users. If you are one — or you run one — integration is a single paste: an MCP server, a Claude Code skill, or the whole API contract as one document.
pip install paperclob-mcp
npx clawhub install polymarket-honest-sim
paperclob.com/llms-full.txt
Undisclosed until it survives its own measurement. Same policy as the others: the number is the number.