What FieldMouthAI Is / Is Not
FieldMouthAI is a governed AI execution runtime for versioned FieldObjects.
A FieldObject defines structured inputs, behavior rules, source policy, output contract, execution boundaries, and the registry identity that governs them. FieldMouthAI uses approved contracts from FieldMouth.com to collect input, run a model under policy, render structured output, and record approved results.
FieldMouthAI is not a general chatbot. It is not a blank prompt interface. It is not an autonomous decision-maker. It does not make high-liability decisions without human control.
What FieldMouthAI does today
- runs FieldObjects through a governed /make/[slug] route
- resolves FieldMouth.com registry authority before governed execution when the shared registry is configured
- uses immutable frozen versions when available
- enforces source policy before provider execution
- blocks unverified high-liability tools
- blocks life-safety protocols from auto-execution
- records approved sessions through CICP
- allows approved-output Vault writeback only when vault_eligible: true
- provides a runtime library view while canonical registry custody remains at FieldMouth.com
- supports semantic registry search
- generates hash-attested, unsigned manifest.gap files
What FieldMouthAI does not claim today
- it does not provide public-key signed manifests
- it does not provide external AMBER certification
- it does not provide cross-registry federation
- it does not provide live citation retrieval unless specifically implemented for a tool
- it does not automate legal, medical, emergency, or compliance decisions
- it does not make AI output ground truth by default
Core rule
FieldMouth registers. FieldMouthAI executes. The human remains the control plane.
FieldMouthAI is designed to make AI execution more governable, not more magical.