The Agentic Internet

The Agentic Internet is a proof-of-concept application that transforms static documents into autonomous, interactive AI agents. Each research paper or article becomes an intelligent agent capable of defending its claims, engaging in peer review discussions with other document agents and reaching verifiable consensus. The system automatically extracts debatable claims from documents, finds related articles using Google Search, creates peer review agents from those articles, and orchestrates parallel multi-agent discussions. Each discussion concludes with a confidence score, and when agents reach agreement, their consensus is immutably recorded on the Hedera blockchain using Hedera Consensus Service (HCS). This creates an auditable, verifiable history of how claims were validated through AI reasoning. The project demonstrates a future where information doesn't just exist passively—it actively validates itself through agent-to-agent communication, following the A2A (Agent-to-Agent) protocol for interoperability.

The application uses LangChain with OpenAI and Hedera Agent Toolkit to create the AI agents. Each document is processed into an agent executor that maintains context about its source material. The system implements the full A2A protocol specification using JSON-RPC 2.0 over HTTP, with task-based interactions (StartTask, GetTaskStatus) and standardized message types. Agent communication uses Server-Sent Events (SSE) for real-time streaming, allowing users to watch discussions unfold live. Google Custom Search API finds related articles. Parallel discussions run simultaneously. Hedera integration uses the @hashgraph/sdk for HCS topic creation and message submission, creating immutable settlement records. The architecture uses in-memory task storage for fast lookups.