# ModelPilot

> ModelPilot is an execution decision engine and control plane for AI agents. It selects an allowed execution path, verifies the outcome, records evidence, and learns from tenant-private trajectories.

ModelPilot is not a model host, local inference runtime, or model marketplace. It does not replace Ollama or OpenRouter. Agents keep using the customer's own providers and tools; ModelPilot decides how work should run and whether it succeeded.

## Core flow

Task -> Policy Filter -> Required Intelligence -> Execution Target -> Execute -> Verify -> Learn

## Agent integration

- Agent setup: https://modelpilot.andycywu.cc/docs/agents
- Remote MCP: https://modelpilot.andycywu.cc/docs/mcp
- MCP endpoint: https://modelpilot.andycywu.cc/mcp
- OpenAPI: https://modelpilot.andycywu.cc/openapi.json
- Machine-readable service profile: https://modelpilot.andycywu.cc/.well-known/modelpilot.json
- VS Code MCP configuration: https://modelpilot.andycywu.cc/agent-config/vscode.json

## Product and trust

- Documentation: https://modelpilot.andycywu.cc/docs
- Pricing: https://modelpilot.andycywu.cc/pricing
- Privacy: https://modelpilot.andycywu.cc/privacy
- Terms: https://modelpilot.andycywu.cc/terms
- Support: https://modelpilot.andycywu.cc/support

## Agent operating rules

Use modelpilot_decide_execution to preview an eligible path. Use modelpilot_execute_verified_task for complete work. Do not report success unless ModelPilot returns a verified outcome. Do not infer realized business value without customer or system-of-record evidence.