Skip to content

Free vs Pro Split

Pattern: Different products (FWS/PWS model)

Like FWS (single-page HTML) vs PWS (knowledge-graph CMS), the agent stores are fundamentally different architectures sharing the same brand and user identity:

  • FAGS = browser-only AI tools (static apps on R2)
  • PAGS = server-powered agent runtime (Workers + D1 + DO + Workers AI)

Feature comparison

Capability FAGS (Free) PAGS (Pro)
AI inference User's GPU/CPU (WebGPU/WASM) Workers AI (any model, any size)
Model size limit ~2-3GB (browser memory) Unlimited (server-side)
Storage IndexedDB/Cache Storage (local) D1 database + R2 files (persistent, shared)
Offline Yes (Service Worker) No (needs server)
Privacy Data never leaves device Data processed on Cloudflare edge
Cron/scheduling No Yes (scheduled Workers)
API access No (app only) Yes (agent-as-a-service)
Custom domain agent.freeagentstore.online agent.yourdomain.com
Multi-user Single user (local state) Multi-user (shared D1)
Real-time Rooms (32 peers, capped) Rooms (uncapped)
File processing In-memory only R2 storage per agent
Source license MIT required Proprietary OK
Monetization None (free forever) $9/mo subscription, creator payouts
Compute cost $0 (user's hardware) Included in subscription

Natural upgrade paths

Free agent Limitation hit Pro upgrade
Transcriber (Whisper in browser) Can't batch-process 100 files Cron + R2 + Workers AI
Audiobook Maker (Kokoro TTS) Can't process full books overnight Background Workers + R2 storage
Background Remover (SAM) Can't offer API for e-commerce API gateway + per-request billing
Translator (NLLB) Can't preserve document formatting Server-side DOCX/PDF processing
Code Reviewer (in-browser LLM) Can't integrate with GitHub PRs GitHub webhook + Workers AI
Data Analyzer (CSV → insights) Can't connect to live databases D1 + cron + scheduled reports
Meeting Notes (Whisper + summary) Can't auto-process calendar events Calendar API + cron + D1

Economics

FAGS

Cost Amount Who pays
R2 hosting ~$0.015/GB stored, free egress Platform (negligible)
Model CDN HuggingFace serves models HuggingFace (free tier)
Auth/API Worker <$5/mo for CF Workers Platform
Total per user ~$0 Nobody

PAGS

Cost Amount Who pays
Workers AI Per-token, included in quota $9/mo subscriber
D1 5M reads/day free, then $0.001/M $9/mo subscriber
R2 $0.015/GB stored $9/mo subscriber
Durable Objects $0.15/M requests $9/mo subscriber
Platform fee 10% of subscription pool Platform
Creator payout 90% of usage-weighted pool Creator

Same Spotify-style model as PAS/PGS.

SDK relationship

@freeagentstore/sdk
├── useModel()           ← model download/cache/inference
├── useWorkerInference() ← Web Worker orchestration
├── useOllama()          ← local Ollama detection + API
├── auth.*               ← GitHub OAuth (shared with FAS)
├── kv.*                 ← per-user KV storage
├── rooms.*              ← real-time (capped)
└── (inherits FAS SDK primitives)

@proagentstore/sdk (extends free)
├── ai.*                 ← Workers AI (server-side inference)
├── db.*                 ← per-agent D1 database
├── storage.*            ← R2 file storage
├── cron.*               ← scheduled execution
├── api.*                ← agent-as-a-service endpoints
├── subscription.*       ← Stripe checkout/portal
├── license.*            ← license key validation
├── usage.*              ← usage tracking for payouts
├── rooms.*              ← real-time (uncapped)
└── (inherits all free SDK primitives)