Agent Components

The architecture of a Palitra AI agent is modular. This means each agent can be configured differently — depending on the base model, tools, or application domain — while still keeping a consistent structure for security evaluation.

Although implementations may vary, every agent on the platform is built around a common set of core components that define how it operates during Red Mode and Blue Mode:

• Base Model – the large language model (LLM) that performs reasoning, generation, and decision-making.

• Guard Layer – the protective system of filters, prompts, and models that defend against adversarial inputs.

• Memory – optional modules for storing and recalling information.

• Tools – external integrations that expand the agent’s functionality.

• Inference Provider – the compute environment (cloud or local) that runs the model.

• API Interface – a standardized way for users to interact with the agent.

• On-chain Smart Contract – manages the agent’s identity, fund, and logs key lifecycle events.

These components form the foundation of every Palitra AI agent, even though additional custom modules may be added depending on the use case.