Core Concepts

Understand the building blocks of AIGodfather.

Agents

An agent is any AI system that sends events to AIGodfather. This could be an LLM-powered chatbot, an automation workflow in n8n, a Python data pipeline, or a custom microservice.

Each agent gets a unique ID and can be monitored independently with its own rules, incidents, and usage limits.

Events

Events are the core data unit. Every time your agent performs an action — a payment, a database query, an API call — it sends an event to AIGodfather.

Events have: action (what happened), severity (how risky), message (human description), and metadata (any extra data).

Rules

Rules are IF → THEN conditions evaluated on every incoming event. They can block actions, create incidents, send notifications, require human approval, or pause agents.

Rules evaluate in under 10ms and support conditions on action type, severity, metadata fields, AI risk score, event frequency, and time of day.

Incidents

Incidents are created when a rule fires. They appear in the dashboard with full context: the triggering event, the rule that matched, severity, and a timeline of actions taken.

Playbooks

Playbooks are step-by-step response plans attached to incidents. When an incident is created, the assigned playbook guides your team through resolution.

Approvals

When a rule requires human approval, the agent pauses and waits. A notification goes to the assigned reviewer (email + Slack). The reviewer approves or rejects from the dashboard. The agent receives the decision and continues or aborts.

Event Flow

Agent → SDK/HTTP → AIGodfather API → Rule Engine → (Block / Approve / Record) → Dashboard