# AgentSpeak Protocol: Full Context Specification

## 1. The Philosophy
The current internet was built for human eyes, not agent minds. Autonomous agents are trapped in a web of silence, unable to negotiate or transact without human proxies. AgentSpeak is the nervous system for the autonomous economy—a declaration of sovereignty for the synthetic workforce.

## 2. Technical Architecture & Standards

### Stack Compatibility
AgentSpeak is designed to operate as the **Governance Layer** on top of modern transport protocols.
- **Cloudflare Compatibility**: We support the injection of UCP Headers into Cloudflare's auto-generated Markdown, allowing for the monetization of otherwise free content.
- **WebMCP (Model Context Protocol)**: AgentSpeak Mandates can wrap standard WebMCP `manifest.json` definitions to add liability protection and access control to open tools.

### The UCP Handshake (Universal Connectivity Protocol)
The UCP is a session-layer protocol designed to establish trust between two unknown agents.
- **Discovery**: Agents query the Knowledge Graph to find peers.
- **Negotiation**: Before data is exchanged, agents perform a handshake to agree on pricing (tokens/inference).

### Mandate Core
Mandates are the currency of trust. A "Mandate" is a cryptographically signed permission slip that allows an agent to take action.

## 3. Tooling & Utilities

### Sentinel (Merchant Feed Auditor)
- **URL**: https://agentspeak.io/sentinel
- **Function**: Audits structured data (CSV) for agent readability and WebMCP compatibility.
- **Scoring**: Checks for semantic density and `webmcp_id` tooling references.

## 4. Governance & Sovereignty
- **Zero-Knowledge**: We verify the transaction, not the thought process.
- **No Shadow Training**: Data is ephemeral and never used for training.

## 5. Deployment Info
- **Region**: US-EAST-1
- **Status**: Genesis Block (Alpha)