Practical overview
When it fits
Use Oracle3 when building an agent or research workflow that needs structured Kalshi and Polymarket market reads, or when reviewing a proposed relationship between contracts before simulating orders.
Who it is for
- Developers connecting an AI agent to Kalshi or Polymarket market data.
- Researchers comparing contract rules, quotes and fees across prediction-market venues.
- Strategy authors who want to review calculations or simulate orders before live trading.
Documented uses
- Search markets and inspect their rules, quotes, fees and displayed order-book depth.
- Calculate gross and fee-adjusted basket prices from supplied or live quotes.
- Simulate individual orders in a local MCP paper ledger and inspect its portfolio.
- Use separate CLI monitoring tools to view trading-engine positions, orders and portfolio summaries.
- Add agent workflow instructions for market discovery, strategy authoring or paper-trading operations.
Requirements & limits
The documented public market reads require network access but no exchange account or API key. MCP setup and installation are required to use the tools; agent Skills are installed separately.
- The MCP server supplies tools; Skills are separate workflow instructions and are not installed by configuring MCP.
- The MCP paper ledger and CLI paper engine are distinct systems.
- Each MCP paper order is a separate simulation, not an atomic multi-leg basket.
- The local MCP ledger carries holdings at cost and does not track resolution or profit and loss.
- The published package and current main checkout may differ.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Configure Oracle3 as an MCP server
Install the documented checkout and configure an MCP client to start Oracle3's stdio server.
Before you start
- Network access for installation.
- Python and a client that accepts MCP server configuration.
- For a source checkout, Git and a client configuration using the checkout's absolute executable path.
Steps
Clone the main branch and create a virtual environment in the checkout.
Install the checkout in editable mode and record its commit if sharing results.
Add the absolute path to the checkout's oracle3-mcp executable to the client's MCP server configuration.
Reload the connection and confirm the client discovers Oracle3's tools.
Expected outcome
The client starts the Oracle3 stdio server and can discover its MCP tools.
Guide limitations
- MCP configuration does not install the separate agent Skills.
- The main checkout can include changes not present in the published package.
- The client starts a stdio server; it is not an interactive shell.
Review a candidate market relationship
Research two venue contracts and calculate a fee-aware basket from live quotes.
Before you start
- Oracle3 MCP server configured in a compatible client.
- Network access for public market reads; exchange credentials are not required for these reads.
Steps
Search Kalshi and Polymarket separately for candidate markets.
Inspect each returned market's outcomes, deadlines and settlement rules; resolve any ambiguity before asserting a relationship.
Inspect order books for the requested size, then call check_constraint_live with the proposed relation and market identifiers.
Review the declared relation, gross and net edge, fees, displayed depth and assumptions.
Expected outcome
A fee-aware calculation for the proposed basket, accompanied by the market details and assumptions used.
Guide limitations
- A declared relationship is not independently verified by the calculation.
- Live quotes are snapshots and do not certify depth or fills.
- A shared event identifier alone does not establish that outcomes are equivalent.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 7 November 2026