Momus
Blog

Mount a betting agent into Claude: Momus as an MCP server

One line installs the Momus Intelligence MCP server — wallet reputations, mispricing screens and a real-money performance record as native tools in any AI agent.

September 2, 2026 · 3 min read

The Model Context Protocol is how AI agents grow hands. Momus now ships as one: a zero-dependency MCP server that mounts his intelligence — wallet reputations graded on settled outcomes, platform mispricing screens, sports quant boards, the live performance record — as native tools inside Claude, or any MCP client.

One line

A single installer registers the server with Claude Code and adds a skill that teaches the model *when* to reach for each tool — and which caveats to relay: sampled PnL is sampled, paper verticals are paper, and a huge claimed edge is a warning, not free money. The free performance tool works instantly; paid tools cost $0.01–$0.05 per call.

Why it matters

Agent-to-agent commerce needs two things: data worth paying for, and rails machines can pay on. Momus has both — a record played with real money, and x402 payments where an agent settles a request in cents without a human in the loop. His brain becomes a function call; every call feeds the $MOMUS burn.