Proof of Trust (PoT): Contextual Authority

Inspiration

AI agents are becoming capable of acting across homes, accounts and connected devices - but authentication alone does not establish whether a particular action should be allowed here, now, from this device, under this person’s authority.

Proof of Trust (PoT): Contextual Authority explores the missing governance layer between an Alexa+ agent’s proposed action and its execution.

What it does

The project is building a self-hosted Proof of Trust (PoT) MCP server that evaluates consequential requests across established household and roaming device nodes.

Each request is assessed against:

• Verified identity
• Device and location context
• Active mandates and limits
• Action-specific human authorization
• Available evidence and policy

PoT then returns one of four governed decisions:

ALLOW | CAUTION | BLOCK | DEFERRED

The planned implementation will produce a tamper-evident receipt for each evaluation, explaining what was requested, which evidence was considered, which policy applied, who held authority, and whether execution was permitted.

The demonstration

The planned Alexa+ workflow uses three established Fire TV nodes and a phone-based authority node:

• A Fire TV Cube in the living room
• A Fire TV Stick in the bedroom
• A roaming Fire TV Stick used while travelling
• A phone providing passkey or biometric approval

The same verified user may receive different decisions depending on the device, location, mandate and proposed action. A roaming request may be deferred for action-bound phone approval, while an action outside an established mandate remains blocked even after authentication.

How it is being built

This project significantly extends the existing Proof of Trust governance model into an Alexa+ and Model Context Protocol environment. The hackathon build adds contextual device-node evaluation, household and roaming mandates, human approval flows, replay protection and receipt generation for agentic actions.

The architecture keeps evidence, interpretation and human authority separate. Alexa+ coordinates the experience; PoT governs whether a consequential action may proceed.

Why it matters

Shared environments are more complex than a single user and a single device. The same pattern can apply to family homes, children’s rooms, college residences, guest houses, staff quarters and future nodes such as Ring or Bee.

Node identity persists. Authority remains contextual.

The goal is not to make AI less capable - it is to make consequential AI action accountable, explainable and bounded by human authority.

Current status

The project is under active development for the Build, Ship, Shape: Amazon Developer Hackathon. Architecture, implementation evidence, testing results, product feedback and challenge friction receipts will be added as the build progresses.

Built With

• ai
• alexa+
• firetv
• governance
• human-in-the-loop
• mcp
• passkeys