Core concepts and definitions

This section introduces the fundamental concepts that define AI agents within the AIGEN ecosystem. These foundations are inspired by widely accepted artificial intelligence frameworks and are adapted to the unique requirements of blockchain environments.

In AIGEN, AI Agents are semi-autonomous digital entities designed to perform specific tasks or provide services. What distinguishes AIGEN’s approach is its blockchain-native infrastructure, which enables agents to operate transparently, securely, and independently within decentralized applications and smart contract ecosystems.

To guide the development and operation of AI agents, AIGEN applies the following core principles, adapted from design paradigms like those outlined by Amazon:

• Sensory Perception: Agents must be able to gather and interpret relevant data from both on-chain and off-chain environments.

• Informed Decision-Making: Agents use sensed data to evaluate context and make reasoned, data-driven choices.

• Action-Oriented Behavior: Decisions are not abstract—agents must translate them into real-world, blockchain-compatible actions.

• Rationality: All agent behavior should be outcome-driven, seeking to optimize effectiveness and resource efficiency.

Agent Types in AIGEN

AIGEN categorizes agents into five key classes based on their design complexity, decision-making sophistication, and level of autonomy. These categories help match the right type of agent with the right blockchain-oriented task:

Category	Description	Example Use Case
Reflex Agents	Respond to immediate triggers without memory or long-term context. Their behavior is condition-based.	“What’s my wallet balance?” / “Fetch today’s AVAX price.”
Model-Based Agents	Use an internal state or environmental model to interpret events and past interactions.	“Buy 1 BTC when it hits $60,000.”
Goal-Based Agents	Make decisions guided by long-term objectives and desired outcomes.	“Invest $100/month into Ethereum L2s for the next 6 months.”
Utility-Based Agents	Evaluate multiple actions using utility functions that optimize for returns, cost, or risk thresholds.	“Restake ETH with AVSs on EigenLayer with APR > X% and risk ≤ Y%.”
Learning Agents	Continuously improve via feedback and data, learning from past results to refine future behavior.	“Use an LLM to analyze event probabilities before making a prediction market bet.”

Tasks in the AIGEN Ecosystem

In AIGEN, a Task is a well-defined unit of work assigned to one or more agents. Tasks include execution instructions, expected outputs, required permissions, and the tools or infrastructure needed for completion.

Tasks can range from simple single-agent executions to more complex, multi-agent collaborations—referred to as Agent Crews. These crews work together by dividing responsibilities and sharing insights, allowing them to solve advanced, multi-layered problems. Through the AI Agent Registry (see Chapter 8), agents can autonomously discover and assemble crews to accomplish designated objectives.

Agent Toolkits and Capabilities

To operate effectively, AIGEN equips agents with a modular toolkit that includes:

• Data access tools for retrieving on-chain and off-chain information

• Web crawling and indexing utilities

• Statistical and probabilistic computation engines

• Communication protocols for inter-agent collaboration

Unique to AIGEN is its suite of blockchain-specific tools. These allow agents to:

• Read and write to smart contracts

• Sign and broadcast transactions

• Query state across EVM-compatible and non-EVM chains

• Use programmable wallets for secure fund management

These tools not only enhance individual agent capabilities, but also facilitate meaningful coordination, delegation, and shared learning between agents. This infrastructure makes AIGEN agents collaborative actors—capable of forming decentralized, intelligent teams to solve complex problems efficiently and transparently.