Agents | Capabilities
The Agent builder enables creators to extend the functionality of their AI Agents by assigning Capabilities — predefined integrations and AI features that determine what the agent can access, perform, or generate.
Capabilities allow an agent to connect with data systems, perform advanced analytics, generate creative outputs, and automate campaign-related tasks across the Athena ecosystem.
When configuring an agent, these options appear under the Capabilities section in the creation form:
Let’s have a look at the available Capability categories and their subfunctions/tools:
Campaign Information
Allows your agent to compile a clear view of your marketing campaigns—by ID, name, or in bulk—so you can quickly see what’s running and where.
Tool | Description |
Campaigns_fetcher | Fetches marketing campaign information. Input: optional campaign_id or campaign_name. If neither is provided, it returns all campaigns. |
Campaign Management
Empowers your agent to build, configure, and test multi-channel campaigns (email/SMS/push), including audiences, schedules, and launch controls on your behalf.
Tool | Description |
Campaigns_creator | Create and configure marketing campaigns (email, SMS/MMS, push).
Supports: Launch types: send now/later/recurring, API-triggered, file-drop triggered, file-based broadcasts Audience/seed lists (segments and/or individual emails) Send-rate limits and per-version volume caps Message templates: email (with confirmed from/reply-to handling), SMS/MMS (provider auto-selection), push notifications Important: Timezone (ZoneInfo format) is required for any scheduling; new configs replace the entire configuration when changing launch types; status changes must only update the status field. |
campaign_proof | Send test messages before launch. Auto-detects recipient types (emails/phone numbers), supports specifying channel and versions. |
ZML Syntax Validator
Enables your agent to safely check Liquid/ZML code to ensure dynamic content and personalization render correctly before any message is sent to your customers.
Tool | Description |
Liquid code validator | Checks Liquid template code for correctness, including supported custom tags, and reports issues |
ZML Snippet
Allows your agent to create, update, and manage reusable HTML content blocks (“snippets”) that can be easily dropped into your templates and campaigns.
Tool | Description |
Create Snippet | Create a new HTML-based “snippet” (content block), with optional metadata, status (active/archived), and editor type |
Update snippet | Update an existing snippet’s HTML/content, metadata, status, or editor. |
Search Snippet | Search for an existing snippet by name |
Data Cloud
Equips your agent to explore and retrieve rich Zeta Data Cloud taxonomies, helping you enhance targeting, insights, and audience definitions.
Tool | Description |
Data_cloud_available_taxonomies | Get the list of available Data Cloud taxonomy categories (e.g., interests, vehicle_ownership, psychographic, b2p, locations, transactions, credit_bureau, demographics_v2, tv_engagement, propensities, zeta_scores). Returns categories only, not entries. |
data_cloud_taxonomies | Search and fetch taxonomy entries from Data Cloud using an enhanced query. Optional filters for taxonomy types and a result limit (default 100, max 1000). |
Data Cloud Insights
Enables your agent to answer natural-language questions about CustomerPulse insights, surfacing who your customers are, what they do, and how best to reach them.
Tool | Description |
Zeta Data Cloud Insights (CustomerPulse) | What it does: Answers questions about CustomerPulse-based insights, for example: acquisition age, gender, income, state segments, online browsing and transactional interests, visitation history, household financial traits, connected TV usage, linear TV ad effectiveness, retention, preferred channels, life events, competitive dynamics, and reachability. Example: “Which state segments show highest retention among high-income millennials?” or “What channels best reach recent movers interested in auto?” |
Experience Management
Allows your agent to design automated customer journeys—defining triggers, delays, splits, and actions—so your lifecycle programs can run on autopilot.
Tool | Description |
Experience Creator | What it does: Creates customer journeys/automations by defining trigger, delay, split, and action nodes (e.g., send a campaign, update profile fields, add to segments). Required inputs: A name for the experience and, optionally, the logic (“expression”)—I can draft that for you based on your rules. Output: Experiences created in the ZMP plus a confirmation message and the full experience details for review (shows the summary unless you want the full JSON). |
Experience Information
Lets your agent search for and retrieve existing experiences/journeys so you can audit, reuse, or refine customer flows without hunting through the UI.
Tool | Description |
experience_fetcher | Fetch a specific “experience” by its ID, or search for experiences by keyword. Options: keyword, sort_by (name, created_at, -created_at, updated_at, -updated_at), per_page. |
Forecasting
Empowers your agent to project performance for ZMP and DSP campaigns based on your budget, timeframe, and targeting so you can plan with data-backed expectations.
Tool | Description |
Forecast Central | What it does: Forecasts campaign efficiency metrics for Zeta Marketing Platform (ZMP: Email/SMS) and Demand Side Platform (DSP: programmatic/RTB).
Key inputs: platform (zmp or dsp), budget, duration (days), optional group_by (platform/channel), optional explode_field (e.g., channel_type, device_type), and targeting filters: demographics (age, gender, income) locations (country/state/city/zip/dma) properties (custom attributes) events (name + time windows and properties) zmp_segments (by segment IDs) devices (DSP device types) dsp_audiences (by audience IDs) brand_safety_categories (IDs) direct_deals (IDs) Important constraints: You can’t mix multiple composed audiences in one request (e.g., an OR across two different multi-condition audience blocks). If a result returns empty, I’ll suggest loosening the criteria. If you want segment/audience inclusion by name, please provide the IDs. Example: “Forecast a 7-day DSP campaign with $15k targeting US iOS users” |
Guidance Model
Equips your agent to surface, manage, and track optimization recommendations and account gaps, helping you continuously improve performance across campaigns
Tool | Description |
Guidance Model – Atomic recommendations | Actions: calculate, list, get_by_id, update_status Use for: generating or managing granular optimization recommendations (by account, campaign, line item, tactic, goal). |
Guidance Model – Recommendation cards | Actions: calculate, get_by_id, get_by_account_ids, update_status
Use for: higher-level recommendation cards based on metrics/goals across one or more accounts. |
Guidance Model – Performance metrics | Action: get_card_metrics
Use for: retrieving performance metrics tied to a specific recommendation card. |
Guidance Model – Account deficiency | Action: get_deficiency Use for: diagnosing account/campaign gaps by source (zmp or dsp), account_id, campaign_id, and goal (revenue, conversion, traffic, awareness). |
Media Campaigns
Enables your agent to create and inspect DSP media campaigns end-to-end—from campaigns and line items to targeting, flights, costs, and inventory objects.
Tool | Description |
Campaign fetcher | Get full details for a campaign by ID. |
Campaign creator | Create a new campaign (label required; optional currency, timezone, roles, goal type, cross-device). |
Line items fetcher | Retrieve all line items under a campaign, including tactics, goals, budgets, pacing, and delivery. |
Flights fetcher | Get flights, their budgets, status, and line item assignments. |
Settings fetcher | Pull frequency caps, viewability, optional settings, and contact assignments. |
Cost constraints fetcher | View margins, external costs, and fee assignments at campaign/line item/tactic levels. |
Targeting fetcher | Inspect all targeting rules (geo, audiences, devices, sites/apps, content categories, schedules, etc.). |
Ads and pixels fetcher | List assigned pixels and all ads tied to line items and tactics. |
Audience search | Find DSP audiences by name/query. |
Direct deals search | Find direct deals by name (excludes disabled/deleted). |
Traffic quality categories | List brand safety categories (optional search). |
Locations search | Find locations (geo) by name/query. |
Model Workbench
Lets your agent discover which ML model targets are available, so it can plug into the right predictive models for your use cases.
Tool | Description |
list_model_target | Lists all available machine learning (ML) model targets from the model workbench registry; returns the complete list of supported model targets/names. No input required. |
Insights
Allows your agent to query pulse reports directly, turning raw analytics into clear answers about performance and audience behavior.
Tool | Description |
pulse_service | Fetches data from a pulse report. Input: a natural-language question and an optional list of verticals. |
Analytics & Reporting
Empowers your agent to run ad-hoc custom performance analysis across campaigns, segments, and dimensions so you can quickly get the metrics that matter most.
Tool | Description |
Marketing report builder | Pulls performance analysis data on your campaigns/segments across dimensions (e.g., campaign, segment, device, subject line, sender, date) with metrics like sends, delivered, opens, clicks, CTR/CTOR, conversions, revenue, bounces, unsubscribes, etc., for any date range and sort order. Example prompts: "Top 20 Email campaigns by revenue in the last 30 days; include conversions and delivered; sort by revenue descending." "Weekly true_open_rate, click_rate, and CTOR by device_type and new_user for last quarter; sort weeks ascending." "For Campaign 12345, compare CTOR, click_rate, and conversions by variant_name and subject since 2025-10-01; only include variants with delivered > 5,000; sort by CTOR descending." |
Segment Information
Enables your agent to understand your data schema and fully manage segments—listing, inspecting, creating, and updating audience definitions as needed.
Tool | Description |
activities_property | Get available activity/event names and their properties (for building event-based filters). |
audience_property | Get available audience properties (demographics, attributes). |
contact_property | Get available contact-related properties. |
resources_schema | Get the schema of available resources (for resource_match conditions). |
segment_lister | Browse/search/list segments with pagination and filters. |
segments_fetcher | Fetch details for a specific segment by ID or name (or all segments if none specified). |
segment_creator | Create a new segment or generate a segment payload (with user/event/campaign/experience rules and prospect/CDP conditions). |
segment_modifier | Update an existing segment’s name, status, or criteria (after fetching the current definition). |
Template Information
Allows your agent to find and retrieve the right creative templates by ID or keyword, streamlining campaign and automation setup
Tool | Description |
Template Finder | What it does: Fetches marketing templates either by exact Template ID or by searching with a keyword (with optional sorting and pagination). Key inputs: template_id (integer) for direct lookup keyword (string) for search order_by: name | created_at | updated_at order_dire: ASC | DESC per_page: integer Examples: “Get template 12345” “Search templates with ‘checkout’, sorted by name ASC, limit 10” |
Web Search & Visit URLs
Empowers your agent to gather external information from the web and specific URLs, supporting research, validation, and richer content generation.
Tool | Description |
Web search | Find current information and sources on the internet (openai_web_search). |
Webpage fetcher/extractor | Retrieve and extract the content of a specific URL to summarize or verify details (url_request) |
Image Generation
Enables your agent to create or edit images from text instructions, giving you on-demand visual assets tailored to your messaging and channels.
Tool | Description |
Image generation (OpenAI gpt-image-1) | Create new images from text or edit existing images. Supports specifying: quality levels (low/medium/high/auto), sizes (square/landscape/portrait), backgrounds (transparent/opaque), and common formats (png/jpeg/webp). |