Skip to main content

Overview

Podium’s Companion API, subscription system, and intelligence layer are designed to support radically different agent experiences on the same infrastructure. These two production apps demonstrate the range — from conversation-first with deep memory to browse-first with creator taste profiles.

Beauty Companion (Sage)

A personal skincare agent: quiz onboarding → conversational discovery → personalized recommendations → checkout → subscription monetization.

User Journey

1

Onboarding quiz

New users complete a 4-step conversational quiz that builds their intent profile. Questions are experiential rather than clinical — “How does your skin feel by early afternoon?” instead of “What is your skin type?”Each answer patches the user’s intent profile and awards points.
2

Conversational discovery

The Conversational Agent uses the profile to deliver personalized recommendations with streaming SSE responses. Users see quick-reply chips for common follow-ups and reason tags explaining why each product was suggested.
3

Purchase

Two purchase modes depending on the product source:
  • Platform products — x402 USDC checkout via concierge orders
  • Affiliate products — redirect to retailer with affiliate tracking
Spend confirmation cards show the product, price, and daily spending limit before the user commits.
4

Subscription monetization

Free users get 25 messages/month and basic recommendations. Sage+ subscribers get unlimited messages and memory-aware personalization — the agent loads accumulated intelligence into every conversation turn.See Subscriptions for the full billing integration.
5

Ongoing engagement

Returning subscribers receive personalized greetings that reference their last conversation. Chat history is durable with full pagination support.

Platform Features Used

Architecture

Key Design: Thin Client

Sage stores almost nothing locally — just a Telegram-to-Podium user mapping. All user data, product data, interactions, orders, and subscription state live in Podium. This makes the companion a thin agentic client that orchestrates the platform API into a coherent experience.

Clone Agents (Familiar)

A creator-based shopping agent: pick a creator → browse their curated products → buy what they’d buy. The agent surfaces products through the lens of a creator’s taste rather than the user’s own profile.

User Journey

1

Creator discovery

Users browse active creator personas via the CreatorPersona discovery endpoint. Each persona has a display name, avatar, platform affiliation, and curated product catalog.
2

Browse creator picks

Selecting a creator loads their resolved product catalog. Products are displayed as swipeable cards with Love, Skip, and Buy actions — each recorded as an interaction.
3

Profile seeding

When a user selects a creator, the creator’s taste profile (preferred brands, price ranges, product concerns) is merged into the user’s intent profile. User-set fields from a completed onboarding quiz are never overwritten.
4

Purchase via affiliate

Buy actions redirect to the product’s affiliate URL (ShopMy, Amazon Associates, or LTK). The platform handles URL resolution and tracking.
5

Guest mode

Unauthenticated users can browse up to 5 products before hitting an auth gate. All interactions during guest mode are replayed to the user’s profile on sign-in — no engagement is lost.

Platform Features Used

Architecture


What This Demonstrates

Both apps share the same Podium infrastructure but deliver radically different experiences: The takeaway: Podium’s primitives are vertical-agnostic and experience-agnostic. The same user model, interaction system, and enrichment pipeline support both a deeply personalized AI companion and a social commerce discovery tool.

Adapting for Your Vertical

To build a companion for a different product domain or experience model:
1

Define your intent profile fields

Replace skin type, concerns, and brand preferences with your domain’s preference dimensions. The profile schema is flexible — use whatever fields describe your user’s preferences.
2

Choose your discovery model

Conversation-first (like Sage), browse-first (like Familiar), or a hybrid. The same API endpoints support all patterns.
3

Curate your product catalog

Populate product records with your inventory, or use CreatorPersona to let creators curate for you. Include images, prices, and external URLs.
4

Pick your monetization

Subscriptions for recurring revenue, affiliate links for commission-based monetization, x402 for direct USDC checkout, or any combination.
5

Deploy to your surfaces

Both apps use the same pattern: one codebase, deployed to web and Telegram. The app detects its surface and renders accordingly.