D
modelpilot_decide_execution
Preview an allowed execution path without running providers.
REMOTE MCP FOR AI AGENTS
Install ModelPilot as an Agent tool layer.
Connect once with a tenant API key. The Agent can then discover policy-aware planning, verified execution, evidence, and value-attribution tools over Streamable HTTP.
MCP endpoint
https://modelpilot.andycywu.cc/mcp
Authorization: Bearer mp_live_...
https://modelpilot.andycywu.cc/mcp
Authorization: Bearer mp_live_...
Available tools
E
modelpilot_execute_verified_task
Execute a complete task and require explicit verification evidence.
G
modelpilot_get_task_evidence
Retrieve the tenant-scoped plan, outcome, cost, and value record.
V
modelpilot_confirm_realized_value
Confirm value only from a customer or verifiable system source.
Automatic-use rule
MCP makes ModelPilot discoverable. Enforced use requires an Agent policy: use modelpilot_execute_verified_task for complete AI work, do not report success unless the result is verified, and do not give the Agent direct provider credentials.