CallPut.appPublic market guide

AI Agents and CallPut MCP

Learn how AI agents can use CallPut MCP to scan markets, evaluate spreads, build unsigned transactions, and track lifecycle state without custodying keys.

Agent Use Cases

Guide

CallPut MCP is meant to give external agents structured market access without giving them custody. A good agent flow helps the user compare markets and prepare a transaction, while keeping approval and signing outside the agent.

  • Discover available markets, expirations, strikes, and option types.
  • Compare quotes, spreads, liquidity, max loss, and payoff constraints.
  • Prepare strategy candidates such as single-leg options or defined-risk spreads.
  • Build unsigned transaction payloads with explicit terms for user review.
  • Track request status, submitted transactions, expiry, and lifecycle state.
  • Explain risks and terms before asking the user to approve anything.

Safe Agent Flow

Guide

Use CallPut MCP as the market and execution-builder layer, then use a wallet or Base Account-compatible flow for user approval, signing, and broadcast.

  • Scan available option chains and market terms.
  • Evaluate spread candidates and risk limits.
  • Build an unsigned transaction payload.
  • Present the transaction for user approval and external signing.
  • Track request keys and lifecycle status after submission.

User Approval Boundary

Guide

The core rule is simple: agents may prepare, but users approve. Any agent or MCP integration should make the final contract, cost, risk, and wallet action visible before signature.

  • The user should see market, strike, expiration, quantity, premium, fees, and max loss before signing.
  • The agent should not hide order terms behind natural-language summaries.
  • The wallet or account flow should be the place where approval and signing happen.
  • A rejected, expired, or failed approval should stop the flow rather than retry silently.

What MCP Does Not Do

Guide

CallPut MCP does not custody funds, hold private keys, bypass user approval, or autonomously sign transactions. Signing and broadcasting stay outside the MCP server.

Agent Safety Checklist

Guide

Agent-facing pages should be strict about refusal and confirmation rules. The safer default is to refuse unclear trades, stale quotes, unsupported markets, or requests that skip review.

  • Refuse unsupported markets, missing strikes, missing expirations, or ambiguous quantities.
  • Refresh quote data when a price, spread, or expiry has become stale.
  • Require explicit user confirmation for strategy, max loss, fees, and settlement terms.
  • Do not present educational examples as investment advice or performance claims.
  • Log transaction intent and returned status so the user can audit what happened.
  • Show recovery paths for failed wallet approval, rejected signatures, and unavailable liquidity.

Integration Surface

Guide

A practical integration should separate market discovery, quote review, transaction construction, user approval, and status tracking. Keeping these steps separate makes failures easier to explain and safer to recover from.

  • Discovery: list supported markets and contract terms.
  • Analysis: evaluate spreads, risk limits, and estimated payoff.
  • Build: create an unsigned payload with the exact terms.
  • Approve: hand off to a wallet or account approval flow.
  • Track: follow submitted status, settlement, expiry, and errors.

Current Scope

Guide

Agent-facing documentation should distinguish the live CallPut product from any narrower public MCP package. Do not imply a wrapper supports every live market unless the specific package documents that support.

FAQ

Can AI agents trade options on CallPut?

AI agents can help scan markets, evaluate spreads, and prepare unsigned transactions. User approval and wallet signing remain required before broadcast.

Does CallPut MCP sign transactions?

No. CallPut MCP does not hold private keys and does not sign transactions. It builds unsigned payloads for external approval and signing.

How do CallPut MCP and Base MCP work together?

CallPut MCP can build the CallPut transaction payload, while Base MCP or a compatible wallet flow can present it for user approval, signing, and broadcast.

Should agents describe SpaceX options as live?

No. SpaceX options should not be described as live unless CallPut officially publishes a live market and documentation.

Can an AI agent choose an options strategy for me?

An agent can help compare mechanics, risks, and candidate structures, but it should not present strategy selection as investment advice. The user remains responsible for the decision and wallet approval.

What should an agent show before wallet approval?

It should show the market, option type, strike, expiration, quantity, premium, mark price, execution price, fees, max loss, max profit, breakeven, and settlement assumptions.

What should happen when an agent sees stale or incomplete data?

The agent should stop, refresh, or ask the user to retry. It should not silently reuse stale quotes, infer missing contract terms, or submit ambiguous transaction payloads.

Open CallPut trading

Review live market terms, max loss, max profit, fees, and wallet approval before opening a position.

Open trading