OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Trading & Crypto Agents

Phil

Phil is an open-source Python AI trading agent for short-term prediction markets. It continuously paper-trades, scores its calibration against market prices, and rewrites its strategy after settled bets. A separate Pearl Connect path supports small capped real-money positions on Polymarket when the operator enables it.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Phil — OpenViST project profile in Trading & Crypto Agents.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Phil when the goal is to study or run an autonomous, continuously evaluated prediction-market strategy, especially with paper trading as the primary feedback loop.

Who it is for

  • Developers studying autonomous AI agents and self-improving strategies
  • Researchers evaluating prediction-market forecasts and calibration
  • Operators interested in a paper-trading workflow with a controlled real-execution path

Documented uses

  • Screen short-term prediction markets for potential differences between an agent estimate and the market price
  • Run continuous simulated trading cycles and evaluate settled outcomes
  • Revise strategy, risk, research, and pacing rules after completed bets
  • Optionally mirror qualifying paper bets with capped real positions through Pearl Connect
Before you choose

Requirements & limits

The repository is public and licensed under Apache-2.0. The documented setup requires Python 3 and Claude Code. Market data uses public Polymarket gamma/CLOB endpoints; the bookmaker-odds benchmark requires an ODDS_API_KEY. The real-execution path requires Pearl Connect and deliberate operator activation.

  • Paper trading is the primary learning engine; real execution is a separate limited path.
  • The agent can rewrite its playbook, risk policy, tooling, sensing, and pacing, while core engine files and caps are protected.
  • The documented experiment excludes crypto up-or-down coin-flip markets.
  • Real execution uses configurable caps and only touches funds through core/real.py and Pearl Connect.
  • The project uses brier_delta as its stated forecast-quality metric rather than relying only on P&L.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run Phil’s documented paper-trading loop

Start repeated paper cycles for short-term prediction-market strategy evaluation.

Before you start

  • Python 3
  • Claude Code available as claude on the PATH
  • Market access through Polymarket’s public gamma/CLOB endpoints

Steps

  1. Run ./loop.sh 10 45 to request 10 paper cycles 45 minutes apart.

  2. Use the command below to produce a calibration and P&L report.

    Documented command
    python3 core/score.py

Expected outcome

Phil runs the documented headless paper-trading cycles and provides a calibration and P&L report.

Guide limitations

  • The project describes this as a research experiment; most trading is simulated.
  • The supplied documentation does not provide further configuration details for the loop.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
30 September 2026
Original website title
bennyjo/phil · Official homepage title · retained in its source language
How we review sourcesSuggest a correction