Cymatic

Cymatic is a voice-first AI agent platform that lets users trigger onchain actions through a normal phone call, without installing an app or using a browser wallet. The project is designed for accessibility, especially for people who do not regularly use smartphone-based Web3 interfaces. A caller’s phone number is mapped to identity context, enabling personalized agent behavior for each user.

The system uses a modular architecture. Gensyn AXL is used for agent-to-agent coordination, KeeperHub Workflows is used for structured and reliable execution, and 0G iNFT is used as the caller-linked onchain identity layer. Together, these components turn voice input into validated intent and execute actions through a controlled, auditable pipeline.

The core value of Cymatic is reducing Web3 complexity at the user layer. Instead of handling wallet extensions, fragmented tools, and multi-step transaction flows, users interact through a simple conversational interface while the backend coordinates identity, orchestration, and execution securely. This makes Cymatic practical for real-world adoption where usability is often the biggest barrier.

We built Cymatic as a modular voice-to-agent pipeline. The backend is in Python (FastAPI), where inbound call/session handling, identity mapping, and agent orchestration are managed. Smart contract components were developed and tested with Foundry, and our iNFT identity contract is deployed on 0G testnet.

The architecture is split into specialized layers instead of a monolith. We use Gensyn AXL for agent-to-agent coordination so voice, reasoning, and execution responsibilities can remain isolated and extensible. We use KeeperHub Workflows for deterministic execution paths, which gave us cleaner operational control and better reliability than ad hoc transaction scripts. We use 0G iNFT as the caller-linked onchain identity primitive, so each phone-number-associated profile can maintain persistent personalization.

A practical hacky decision we made was to support graceful degradation across data sources and infrastructure states during development. We separated user-facing flow from strict dependency coupling so pieces could continue functioning even when one upstream source was unstable. That helped us iterate quickly while still validating end-to-end identity + execution behavior. Overall, partner technologies gave us speed (KeeperHub), modularity (AXL), and verifiable identity continuity (0G iNFT)