| Model | LLM provider (OpenAI, Anthropic, Google, OpenRouter), specific model, temperature, and effort |
| Internal name | The agent’s identifier within the platform — visible to the team, not to the customer |
| Role | The agent’s function and persona — the “position” it holds in the operation |
| Prompt | System instructions with AI editor, version history, and real-time quality analysis |
| Temperature | Degree of creativity vs. consistency of responses (0.00 to 1.00) |
| Effort | Reasoning depth controlled by maximum number of steps (10 to 100) |
| Capabilities | Optional features: planning, reasoning, automatic date and time |
| Conditional prompts | Instruction blocks activated by conversation or contact variables |
| Internal chat | Direct testing interface with the agent without going through any external channel |
