​
General Concepts

  1. AI Agent (also know as Agents): Agents are a new automation capability within the Blink platform that enables AI reasoning (human-like thinking). Agent reasoning involves goal-directed decision-making that can resemble human-like thinking, but is based on algorithms and data rather than consciousness.
  2. No-Code Agent Builder: Blink’s visual interface for creating and configuring AI agents without writing code. It allows users to define the agent’s purpose, set abilities and constraints, customize the prompt template, and manage interactions using simple English instructions (e.g., “You are a GRC analyst responsible for handling policy violations and escalating critical compliance risks”), all within an intuitive UI.
  3. LLM: A Large Language Model (LLM) is an AI system trained on large amounts of text to understand and generate human language. It can answer questions, write text, summarize information, and more.

​
Agent Configuration Concepts

  1. Agent Roles: User-defined scopes that describe what the agent is responsible for. Roles help shape the agent’s behavior and context (e.g., “Customer Support Agent,” “Security Analyst Assistant”).
  2. Agent Constraints: Rules that define what an agent should not do. These boundaries limit the agent’s behavior, inputs, or outputs to ensure it operates safely, ethically, and within the intended scope (e.g., “Do not assume intent. Only report what you can observe from the logs” or “Only respond with three sentences”).
  3. Agent Abilities: The specific actions an agent can perform, implemented via deterministic and secure Blink workflows (e.g., “Hybrid Analysis - Hash Report” or “VirusTotal - IP Address Report” ).
  4. Agent Knowledge: Consists of uploaded files and documents, that the agent can reference to enhance its understanding of your environment. These resources enable the agent to provide more accurate responses and make informed, context-aware decisions.
  5. Peer Agents: Other Blink AI agents that can delegate sub-tasks. They may pass tasks between each other, share context, or operate in sequence to complete multi-step processes across teams or systems.

​
Agent Run-Time Concepts

  1. Agent Session - In a session an agent performs a single task or a series of related tasks to complete a specific objective. Agents operate within isolated sessions and begin each one without any memory of previous sessions. During a session, they gather, learn and process information relevant to the task at hand.
  2. Agent Task: A specific goal or instruction given by user to the agent to complete. The Agent tasks are defined in the agent’s workflow action and work in conjunction with agent roles and constraints
  3. Agent Reasoning: Agent reasoning refers to the interaction between the agent and the large language model (LLM). It represents the internal reasoning, planning, or decision-making process the agent performs before taking an action.
  4. Data Processed: Data processed refers to the information ingested, analyzed, or transformed by the agent during its task. This includes but not limited to agent configuration, any inputs from user prompts or outputs generated during workflow execution
  5. Agent Operation - An agent operation refers to a single, discrete execution carried out by the agent, such as running an ability (executing a workflow) or delegating a sub-task to a peer agent.
  6. Agent Output: The agent’s final output for a given task. The format can be either free text or JSON, depending on the instructions provided.