On this page

Automation

Tajo provides three levels of automation for customer engagement, from simple rule-based workflows to fully autonomous AI agents.

Three Layers of Automation

1. Skills, Deterministic Workflows

Skills are atomic, rule-based automation units. They execute the same way every time: a trigger fires, conditions are checked, actions run.

triggers:
  - event: cart_abandoned
    conditions:
      - cart_value: "> 50"
      - time_since_activity: "> 30m"

actions:
  - brevo/send-email:
      template_id: 101
      to: "{{ contact.email }}"

Best for: Repeatable processes, cart recovery, welcome sequences, data sync, order confirmations.

Browse Skills | Skills Format

2. Agents, AI-Powered Orchestration

Agents use LLMs to reason about goals, select tools, and adapt strategy. They compose multiple Skills and Brevo MCP tools to achieve complex marketing objectives.

---
name: win-back-agent
tools:
  - brevo_contacts
  - brevo_segments
  - brevo_email_campaign_management
  - brevo_sms_campaigns
---

# Win-Back Agent

Analyze churned customers, design personalized re-engagement
sequences based on purchase history, and execute across
the optimal channel mix.

Best for: Complex decisions, choosing channels, personalizing offers, adapting sequences based on customer behavior.

Build an Agent | Agent Specification

3. Natural Language, Intent-Driven Marketing

The top layer translates marketer intent into agent actions:

“Create a VIP appreciation campaign for customers who spent over $500 this quarter. Use email with a personal thank-you and an exclusive 20% discount code.”

The orchestration layer selects the right agent, which composes the right skills and Brevo MCP tools to execute.

MCP Architecture

Choosing the Right Level

Question	Use Skills	Use Agents
Is the logic the same every time?	Yes
Does it require judgment calls?		Yes
Single trigger → single action?	Yes
Multi-step with branching logic?		Yes
Needs to adapt to customer behavior?		Yes
High-volume, latency-sensitive?	Yes
Needs human-readable audit trail?	Both	Both

How They Work Together

In practice, agents invoke skills as part of their execution:

Marketer: "Set up cart recovery for the holiday season"
    ↓
Agent: Reads context, decides on strategy
    ↓
Agent calls Skill: tajo/recover-abandoned-cart (configured for holidays)
Agent calls Skill: tajo/customer-sync (ensure data is fresh)
Agent calls MCP: brevo_segments (create holiday shopper segment)
Agent calls MCP: brevo_templates (check holiday email templates exist)
    ↓
Result: Cart recovery active with holiday-specific messaging

Triggers

Both Skills and Agents support three trigger types:

Event Triggers

React to something happening in your system:

triggers:
  - event: order_completed
  - event: customer_created
  - event: cart_abandoned

Schedule Triggers

Run on a recurring schedule:

triggers:
  - schedule: "0 9 * * MON"    # Every Monday at 9am
  - schedule: "0 */4 * * *"    # Every 4 hours

Webhook Triggers

Invoke via HTTP request:

triggers:
  - webhook: /automation/cart-recovery/trigger
    method: POST

Brevo MCP Integration

All automation runs through Brevo’s official MCP server, which exposes 27 modules as AI-callable tools. Tajo agents connect to the specific modules they need:

Module	What It Does
`brevo_contacts`	Manage contacts and lists
`brevo_email_campaign_management`	Create and send email campaigns
`brevo_sms_campaigns`	SMS campaigns
`brevo_whatsapp_campaigns`	WhatsApp campaigns
`brevo_segments`	Dynamic contact segments
`brevo_campaign_analytics`	Campaign performance data
`brevo_deals`	CRM deal management

Full server list

Next Steps

Brevo MCP Server Setup, Connect Brevo’s 27 MCP modules
Building Your First Agent, Hands-on tutorial
Skills Reference, Browse available automation skills
Agent Specification, Define custom agents