AI Agent Configuration (timezone, format, options)

The AI Agent has several configuration options grouped in the edit panel (Management → AI Agents → Edit). This guide covers the most important ones, starting with regional configuration (timezone and date format) which directly affects how the agent responds about dates and times.

🌍 Regional configuration (timezone and date format)

Each AI Agent has two selectors that define how it understands and communicates time:

You'll find them in Basic Configuration when editing the agent:

Basic Configuration
├── Agent Name
├── Timezone        ← here
├── Date format     ← here
└── Description

🏢 Hierarchy: organization, workspace, agent

MuseRelay has a hierarchical structure:

Organization
├── Workspace 1
│   ├── Agent A
│   └── Agent B
└── Workspace 2
    └── Agent C

Regional configuration lives at two levels:

1. Organization level (all agents inherit by default) — set under Settings → Organization.
2. Agent level (optional override) — the agent can have its own timezone/format that overrides the organization.

Resolution cascade

For each agent, the system resolves the configuration in this order:

1. Does the agent have an explicit value? → use that.
2. If set to "Inherit" → does the organization have a value? → use the organization's.
3. If both are empty → fallback:
   • Timezone → UTC.
   • Date format → en_US.

Practical example

A retail chain with organization "Screen Art" configured as Europe/Madrid + es_ES:

🕐 What changes with the timezone

The agent's timezone affects all _local variables in the prompt and the auto-injected context block. Example with the same real time (April 17, 2026, 15:47 UTC):

📦 Auto-injected context block

Regardless of whether you use variables in your prompt, the system always prepends to the start of your agent's prompt this block before sending it to the model:

## Current context (injected by the system, do not modify)
- Current time: 15:47 (Europe/Madrid)
- Date: Friday, April 17, 2026
- Tomorrow: Saturday, April 18, 2026
- Agent timezone: Europe/Madrid
- Format locale: es_ES

This ensures the agent always knows the real date and time, without depending on you remembering to use variables.

💡 Recommendations

• By default, leave the agent selectors as "Inherit" and configure the zone at the organization level. This way all agents use the same configuration and you only change one place.
• Override only when needed: if the same organization serves multiple countries, create agents with per-zone overrides (one agent per country).
• For international audiences where customer time matters, consider creating one organization per country — the cleanest pattern in multi-tenant.
• Hot changes: if you edit the timezone of the agent or the organization, the model sees the new values in the next conversation. No restart needed.

🛠️ Other agent configuration options

Besides the regional config, the agent edit panel has:

• Name and description — what your agent is called and what it does.
• Prompt configuration — the system prompt that defines the behavior. See Prompt variables.
• Custom variables — your own variables in JSON (e.g. {{company}}, {{store_hours}}).
• Calendar Commands (Cal.com) — so the agent can book appointments. Requires API key + Event Type ID.
• Advanced configuration — streaming, max tokens, creativity (temperature).

Each one is documented separately in this section.

Frequently asked questions

🕐 Who sees what time? Embed chatbot vs agent panel

A common question: who exactly is affected by the regional settings of the agent? The answer matters because the AI agent and the humans viewing conversations may be in different timezones.

Two interfaces, two different rules

Why the embed chatbot uses the browser's timezone

If a Mexico City customer reads the chat on their web:

• Their clock shows 09:15.
• If the bot shows "I'll reply at 09:15" → matches their clock ✅.
• If it showed "I'll reply at 16:15" (Madrid time) → confusion ❌.

It's what WhatsApp, Messenger, Telegram, Slack do: everyone sees the time in their own zone. Zero configuration, zero confusion for the end customer.

Why the agent panel uses the agent/operator's timezone

The MuseRelay human operator is working and needs to see times in their work zone. If they handle support from Madrid but a Mexican customer wrote at 02:15 CDMX (09:15 Madrid), the operator wants to see "09:15 Madrid" to know when to reply.

In the internal panel the system uses this cascade:

1. Operator user's timezone (if set).
2. Organization's timezone.
3. Fallback to Europe/Madrid.

Modern chat-style day separators

The embed chatbot groups messages by day using the visitor's browser timezone, like WhatsApp or Telegram:

─────── Today ───────
14:30  Message 1
14:32  Message 2

─────── Yesterday ───────
09:15  Message 3

─── Wednesday, April 15 ───
11:00  Message 4

─── March 2, 2026 ───
Older message

Separator rules:

• "Today" → messages of the visitor's current day.
• "Yesterday" → previous day.
• Weekday + date → last 7 days.
• Full date → older.

Each individual message shows only the time (14:30). The separator already says what day, so it doesn't repeat.

Visual summary

Each side sees the useful time without stepping on the others.