Overview

The Agents screen manages AI agent configurations — the LLM-powered automations that run on the platform (match advisor, conversation agent, profile analysis, and others). Each agent is stored as a General Code of type Agent.

The Agents page is a purpose-built editor for agent General Codes. It exposes all editable fields in a structured form and handles encrypted API key storage. Changes take effect on the next invocation of the agent — no restart or deployment is required.

Agent list and detail are at Settings → Agents.

Agent fields

FieldWhat it controlsWho can edit
Name Unique identifier for the agent. Used as the lookup key in code. Editable by SManager; changing this key after deployment requires a code update. SManager only
Description Plain-language note for your team — what this agent does. Not sent to the LLM. SManager only
Model LLM model name passed to the API (free-text, e.g. gpt-4o). Must match a model ID supported by the configured API key. SManager only
Tokens Maximum output token limit per request. Higher limits allow longer responses but increase cost per call. SManager only
Temperature Randomness of LLM output (numeric). 0 = deterministic, higher values = more varied output. Most agents use 0–0.5. SManager only
API Key API key for the LLM provider. Stored AES-encrypted in the database; displayed masked in the UI. Enter a new value and save to rotate the key. SManager only
Prompt The base system prompt sent to the LLM. Defines the agent's role, rules, and tone. Changes here are high-impact — test on a staging environment before applying to production. SManager only
Extra Instructions Brand- or workflow-specific additions appended to the base prompt on save. Separated from the base prompt by a --- divider so the two sections are kept distinct. Regular editors can update this field without SManager access. Edit role (any admin with edit permission)

The agent also has an Active / Inactive status switch. Inactive agents are disabled and will not be invoked by the platform. SManager can toggle this status.

Prompt vs Extra Instructions

The agent prompt is stored as two parts:

  • Prompt — the stable base written by engineering or product. SManager-only to protect core agent behavior.
  • Extra Instructions — brand-specific or campaign-specific additions that operators can adjust without touching the base prompt. Any user with the Edit role can update this field.

When the agent is invoked, both sections are joined into one combined prompt with a --- separator between them. When you load an existing agent, the editor splits the stored prompt back into these two fields automatically.

This separation means you can safely give content or customer-success teams the ability to tune agent behavior (Extra Instructions) without risking accidental changes to the core prompt logic.

Permissions

ActionRequired permission
View agent listView access to General Codes Settings page
Edit Prompt, Model, Tokens, Temperature, API KeySManager on General Codes Settings page
Edit Extra InstructionsEdit on General Codes Settings page
Toggle Active/InactiveSManager on General Codes Settings page
Create new agentSManager on General Codes Settings page

Creating and editing

1
Open Agents
Go to Settings → Agents. The list shows all configured agents with their name and active status.
2
Open an agent or create new
Click a row to edit an existing agent, or click Add Agent to create one.
3
Edit fields
Update the fields you have access to. For Extra Instructions: enter brand-specific additions without modifying the base Prompt. For Model: enter the exact model ID as expected by the API (e.g. gpt-4o).
4
Save
Click Save. The Prompt and Extra Instructions are joined on save. The new configuration is used on the next agent invocation. No restart needed.