# Oracle3 prediction-market arbitrage checks [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/YichengYang-Ethan/oracle3-prediction-market-agent  
**GitHub Stars:** 443  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/oracle3-prediction-market-arbitrage-checks

## Description
Kalshi and Polymarket data, venue fees, no-arbitrage checks and paper execution for AI agents.

## Claude Desktop Quick Installation
Heuristic fallback — verify the package name and runner against the repository README before running it. Uses `npx` (confidence: low):

```json
"mcpServers": {
  "oracle3-prediction-market-arbitrage-checks": {
    "command": "npx",
    "args": ["-y","oracle3-prediction-market-arbitrage-checks"]
  }
}
```

## Documentation

## What Oracle3 prediction-market arbitrage checks does

Oracle3 prediction-market arbitrage checks provides an MCP interface for researching prediction markets on Kalshi and Polymarket. Agents can search contracts, retrieve market metadata and resolution rules, inspect quotes and displayed order-book depth, compare related outcomes, estimate fees, and record simulated trades in a local paper ledger.

The server is aimed at workflows where an agent needs structured market inputs before evaluating a strategy. Its relationship analysis covers implication, exclusivity, complement, same-event, and event-sum cases. Results include candidate baskets, per-leg fees, calculated gross and net edge, and the assumptions used by the check.

## How it works

After installation, an MCP client starts the `oracle3-mcp` executable over the local process interface. Public Kalshi and Polymarket reads do not require exchange credentials. The agent can combine discovery and inspection tools with offline calculations, or use current venue data when analyzing live market conditions.

The named tools include `search_markets` and `get_market` for discovery; `get_quote` and `get_orderbook` for market conditions; `check_constraint` and `check_constraint_live` for relationship analysis; `trading_fee` and `fair_value` for calculations; and `paper_order` and `paper_portfolio` for simulated fills and holdings. The README identifies 13 MCP tools in total.

Oracle3 prediction-market arbitrage checks calculates edges from supplied relationships, prices, quantities, and fee schedules. A result is conditional on the relationship and quotes being correct; the calculation does not by itself establish that two contracts resolve identically or that an opportunity can be executed.

## Setup and configuration

The documented setup uses Python 3.10 or newer. Clone the repository, create a virtual environment, and install the checkout with `pip install -e .`. Configure an MCP client with the absolute path to `.venv/bin/oracle3-mcp` and no command arguments. Windows installations use the corresponding files under `.venv\\Scripts`.

The README gives a Claude Code command that registers the server from the project directory. It also states that other MCP clients can use the executable path as the command with an empty argument list. No exchange credentials are needed for public market reads or the local paper ledger.

## Tools and capabilities

- Search Kalshi or Polymarket markets and retrieve contract details.
- Read quotes, fee information, order books, and displayed depth.
- Check five logical relationships between contracts.
- Calculate trading fees and parameterized fair-value outputs.
- Simulate orders and inspect local paper holdings.
- Build further strategies with the repository's Python interfaces and separate CLI engine.

Oracle3 prediction-market arbitrage checks also ships six reusable agent workflows for data discovery, constraint arbitrage, quant strategy authoring, agent strategy authoring, paper operations, and live operations. These Skills are separate workflow guidance layered on top of the MCP tools.

## Limitations and notes

Oracle3 prediction-market arbitrage checks exposes market data, calculations, and paper operations through MCP. The README separates real-order execution from this interface: live trading uses the project's Python and CLI engine rather than the MCP paper tools. The engine includes paper and live venue adapters, but using those adapters is outside the documented MCP quickstart.

Calculated net edge includes the selected venue fee schedule, but it remains a model result based on the supplied prices, quantities, and relationship assumptions. Agents should inspect resolution rules and contract definitions before treating markets as comparable. Optional venue, data-provider, and wallet integrations are maintained in the separate `oracle3-extras` project.

_Full upstream README: https://allmcps.com/mcp/oracle3-prediction-market-arbitrage-checks/readme_