Overview
On-Chain Agents are defined as programmable externally-owned accounts (EoAs) that are spun up via an agent provider’s CLI. These agents receive instructions either through a fixed python script or an off-chain ML model, resulting in two distinct workflows.Static
Static agents operate based on non-upgradable instructions. These agents follow a fixed path, interacting exclusively with whitelisted smart contracts. They subscribe to specific events, reacting to each event as predefined.
Deterministic Approach for Agentic Workflow
Dynamic
Dynamic agents operate autonomously, receiving instructions from an off-chain ML model. The model’s inference output is first verified and proved on-chain before the agent acts. These agents adapt based on the data and parameters set within the ML model, allowing for evolving, non-fixed pathways.
Non-Deterministic Approach for Agentic Workflow