Skip to content

Oracle3

Build prediction-market agents that discover markets, analyze opportunities, and test strategies.

Connect agents to Kalshi and Polymarket through MCP, guide their work with reusable Skills, and develop strategies with a Python execution engine.

13 MCP tools · 6 agent Skills · Kalshi & Polymarket data · Python 3.10+

Agent quickstart · MCP tool reference · Project and contributions · GitHub

Explore the toolkit

Build What Oracle3 provides
Market research agents Contract discovery, resolution rules, quotes, and order books
Related-contract analysis Five relationship checks, shared fees, and structured basket calculations
Paper-trading workflows Simulated entries and local holdings through MCP
Python strategies Strategy interfaces, risk controls, and paper/live venue adapters
Agent experiments Reusable Skills, observable state, and reproducible software validation

Public market reads need no exchange credentials. MCP provides data, calculations, and a local paper ledger. Real-order execution uses the separate CLI/Python engine.

Connect your agent

Install the current development checkout:

git clone https://github.com/YichengYang-Ethan/oracle3-prediction-market-agent.git
cd oracle3-prediction-market-agent
python -m venv .venv
.venv/bin/python -m pip install -e .
claude mcp add oracle3 -- "$PWD/.venv/bin/oracle3-mcp"

For other MCP clients and Windows setup, follow the agent quickstart. Published-package installation is documented in the MCP reference.

Then ask your agent:

Find Kalshi and Polymarket contracts about the next Fed rate decision. Compare their outcome definitions, resolution rules, and order-book quotes. Summarize the markets with their IDs, prices, and comparison assumptions.

Prefer an offline first task? Try the worked fee-aware calculation.

Tools, Skills, and an execution environment

Use individual MCP tools for structured data and calculations, or load one of the six agent Skills for a complete workflow. The Python interfaces support strategy development and engine operations; oracle3-extras hosts optional data, venue, and wallet integrations.

For agent developers, llms.txt provides a machine-readable guide to tasks, tools, and project resources.

From U Lab research to an open-source project

Yicheng Yang worked with Haofei Yu on Coinjure at U Lab and co-developed the original U Lab Oracle3 codebase. He continues to extend Oracle3 with MCP integration, reusable agent workflows, and infrastructure for experimentation.

Read the project overview and contributions for the background, current capabilities, and direction for agent research.

Go deeper