WonderTwin

A local development tool that runs stateful, behavioral simulations of external services like Stripe, Plaid, and Shopify, enabling AI agents to develop and test against realistic service behavior without mocks or vendor sandboxes. It targets developers and engineering teams at mid-market and enterprise companies who build AI agents that depend on third-party APIs. The tool is delivered as a single Go binary that runs on a developer's laptop or CI runner, with a registry of open-source and pro twins, and is continuously calibrated against production behavior for pro versions.

Key features

• Local twins of external services
• Stateful behavioral models
• Drift-aligned calibration
• Production-fidelity rate limits
• Deterministic replay
• MCP-compatible discovery
• CLI and MCP support
• Open source and pro twins