WHAT DOES AN AGENT CONSIST OF?

An AI Agent is not a new model.

It is a model placed inside a specific working structure.

On F68 AI, users configure an Agent through three main layers:

SYSTEM PROMPT

The System Prompt shapes how the Agent must work.

It can include:

• the task;
• the scope;
• the principles;
• how to use sources;
• the stopping conditions;
• the output requirements.

System Prompt = how the AI must work.

CONTEXT

Context holds what the Agent needs to know to carry out the task.

It can include:

• domain knowledge;
• documents;
• background data;
• methods;
• standards;
• examples;
• experience that has been systematized.

Context = what the AI needs to know to do the job right.

TOOL

A Tool lets the Agent fetch more data or do things the model cannot do on its own with the existing Context alone.

For example:

• web search;
• reading files;
• querying data;
• running code;
• calculating;
• calling APIs or integrated systems.

Tool = a capability the Agent can use when the job requires it.

Workflow, rules, output schema, source policy, or failure behavior can live in the System Prompt or in the runtime layer that is deployed.

Specialized data can live in the Context or be fetched through a Tool.

The user-facing structure:

Model + System Prompt + Context + Tool → AI Agent

At the operating level, F68 AI also uses Runtime/Orchestration to assemble Context, call Tools, and manage the Agent lifecycle.

Runtime is not a separate configuration field in the interface.