Normalized for Mintlify from knowledge-base/neurigraph-memory-architecture/legacy-brain-api-pricing-strategy.mdx.
aiConnected Brain API
Strategic Product Document
Executive Summary
Brain is a persistent memory layer that sits above all AI platforms. Users maintain continuous context across ChatGPT, Claude, Gemini, and any MCP-compatible service. Brain is the foundation of aiConnected’s cognitive operating system for AI and future robotics.
The name earns its meaning: aiConnected = All AIs connected through persistent memory.
The Problem
Every person using AI today has fragmented conversations:
- ChatGPT knows about your project but Claude doesn’t
- You explained your preferences to Gemini last week, now you’re starting over with GPT
- Switch models and lose everything
- Even within the same platform, context disappears after the conversation ends
The major platforms (OpenAI, Anthropic, Google) will never build cross-platform memory. Walled gardens serve their business interests. That interoperability gap is aiConnected’s entire market.
The Solution
Brain is a persistent memory layer that:
- Stores conversation context across all AI platforms
- Enables continuous memory regardless of which AI you use
- Learns and accumulates knowledge over time
- Creates switching costs that compound with usage
Target Markets
Consumer Market
- Power users juggling multiple AI platforms
- Developers and creators needing continuity
- Professionals wanting AI that learns their preferences
- Teams needing shared context
- Marketing agencies deploying AI for clients
- Service businesses using AI-powered chat and voice
- Enterprises wanting persistent AI relationships with customers
Pricing Structure
Consumer Tiers
| Tier | Base Price | Includes | Overage |
|---|
| Starter | $1/mo | 10 free actions | $0.50/action |
| Personal | $9/mo | 100 actions | $0.25/action |
| Pro | $19/mo | Unlimited actions | - |
| Teams | $39/seat | Unlimited + shared memory | - |
Storage Add-On
| Add-On | Price | Includes |
|---|
| Storage expansion | $5/mo | +100,000 memories |
Included Storage by Tier
| Tier | Included Memories |
|---|
| Starter | 1,000 |
| Personal | 10,000 |
| Pro | 100,000 |
Conversion Economics
The pricing structure drives natural upgrades:
| Monthly Actions | Starter Cost | Personal Cost | Pro Cost |
|---|
| 10 | $1 (free actions) | $9 | $19 |
| 15 | $3.50 | $9 | $19 |
| 20 | $6 | $9 | $19 |
| 25 | $8.50 | $9 ✓ | $19 |
| 50 | $21 | $14 | $19 ✓ |
| 100 | $46 | $34 | $19 ✓ |
- At 25 actions/month, Personal becomes the better value
- At 50 actions/month, Pro becomes the better value
- The $1 entry point filters out non-serious users while remaining accessible
- The 10 free actions (a $5 value) let users experience the product before paying overage
Revenue Projections
Distribution Assumptions
| Segment | Percentage | Average Monthly Revenue |
|---|
| Starter | 50% | 4(1 + ~6 paid actions) |
| Personal | 30% | 11(9 + overage) |
| Pro | 15% | $19 |
| Teams | 5% | $39 |
Brain Revenue by Scale
| Users | Monthly Revenue | Annual Revenue |
|---|
| 1,000 | $10,100 | $121,200 |
| 10,000 | $101,000 | $1,212,000 |
| 100,000 | $1,010,000 | $12,120,000 |
| 1,000,000 | $10,100,000 | $121,200,000 |
Personas Revenue (Additional)
Average 1.5 personas per user at $9/month average:
| Users | Persona Revenue/Month | Persona Revenue/Year |
|---|
| 1,000 | $13,500 | $162,000 |
| 10,000 | $135,000 | $1,620,000 |
| 100,000 | $1,350,000 | $16,200,000 |
| 1,000,000 | $13,500,000 | $162,000,000 |
Storage Upsell Revenue
Assuming 10% of users purchase average 2 storage add-ons:
| Users | Storage Revenue/Month | Storage Revenue/Year |
|---|
| 100,000 | $100,000 | $1,200,000 |
| 1,000,000 | $1,000,000 | $12,000,000 |
Combined Revenue (Brain + Personas + Storage)
| Users | Monthly Revenue | Annual Revenue |
|---|
| 1,000 | $23,600 | $283,200 |
| 10,000 | $236,000 | $2,832,000 |
| 100,000 | $2,460,000 | $29,520,000 |
| 1,000,000 | $24,600,000 | $295,200,000 |
Infrastructure Costs
Compute Strategy
Self-hosted LLM (Ollama or similar) rather than paying per-token to third-party providers.
Rationale:
- Summarization and keyword extraction don’t require frontier models
- Chinese models (DeepSeek, Qwen) cost ~$0.01 per 1M tokens
- Self-hosted reduces this to fixed infrastructure cost
- Break-even vs API pricing occurs around 200-400 active businesses
Compute Costs by Scale
| Users | Monthly Infrastructure |
|---|
| 1,000 | $3,000 |
| 10,000 | $8,000 |
| 100,000 | $50,000 |
| 1,000,000 | $150,000 |
Storage Architecture
| State | What’s Stored | Size |
|---|
| Active (recent) | Full conversation, uncompressed | ~2MB |
| Archived (after X days) | Full conversation, compressed | ~200KB (90% compression) |
| Search index | Summary + keywords + embeddings (internal only) | ~10KB |
Storage Costs at 100,000 Users
Assuming 10,000 memories average per user, 90% archived:
| Type | Size | Cost/GB/Month | Monthly Cost |
|---|
| Hot (active + index) | 210 TB | $0.02 | $4,200 |
| Cold (archived) | 180 TB | $0.004 | $720 |
| Total | 390 TB | | $4,920 |
Storage Costs at Scale
| Users | Total Storage | Monthly Cost |
|---|
| 100,000 | ~400 TB | $5,000 |
| 1,000,000 | ~4 PB | $50,000 |
Profit Projections
At 100,000 Users
| Item | Monthly |
|---|
| Revenue (Brain + Personas + Storage) | $2,460,000 |
| Compute | $50,000 |
| Storage | $5,000 |
| Total Cost | $55,000 |
| Profit | $2,405,000 |
| Annual Profit | $28,860,000 |
At 1,000,000 Users
| Item | Monthly |
|---|
| Revenue (Brain + Personas + Storage) | $24,600,000 |
| Compute | $150,000 |
| Storage | $50,000 |
| Total Cost | $200,000 |
| Profit | $24,400,000 |
| Annual Profit | $292,800,000 |
Gross Margins
| Scale | Gross Margin |
|---|
| 1,000 users | ~70% |
| 100,000 users | ~98% |
| 1,000,000 users | ~99% |
Technical Architecture
Integration Model
Brain operates as an MCP (Model Context Protocol) server:
- Integrates the same way as Google Calendar, Gmail, GitHub MCPs
- Works with Claude natively
- Other platforms integrate as MCP adoption grows
- No proprietary protocol needed; uses existing standard
Memory Operations
| Operation | What Happens |
|---|
| Store | Full conversation saved, compressed after X days, summary and keywords generated internally for search |
| Search | Internal summaries and keywords searched, relevant memories identified |
| Retrieve | Full conversation decompressed and returned to active context |
- Memory exists
- Token count
- Attachments
- Timestamp
- Source conversation
What Stays Hidden (Trade Secret)
- Summaries (internal indexing only)
- Keywords (internal indexing only)
- Ranking/relevance algorithm
- Compression method
- Retrieval logic
- Storage architecture
Competitive Protection
Strategy: Trade Secret Over Patent
| Approach | Trade-Off |
|---|
| Patent | Requires full public disclosure; 20-year protection in theory |
| Trade Secret | No disclosure; protection lasts as long as secret is kept |
- Enforcement against well-funded competitors costs millions
- International coverage is limited (China doesn’t honor US patents)
- Competitors can design around patents
- Software patents are increasingly weak in courts
- Filing requires disclosing exactly what we’re protecting
Protection Layers
| Component | Strategy |
|---|
| Core memory mechanism | Trade secret |
| Brand (aiConnected, Brain) | Trademark |
| Code | Copyright (automatic) |
| UI/UX innovations | Possibly patent if unique enough |
What Protects Us Over Time
| Timeframe | Primary Moat |
|---|
| Year 1 | Secrecy + speed to market |
| Year 2 | User data gravity (memories accumulate) |
| Year 3+ | Ecosystem lock-in (personas, integrations, agencies) |
Why Competitors Can’t Easily Replicate
| Factor | Protection |
|---|
| No public spec | 6-12 months of guessing for competitors |
| Hidden summarization logic | They build inferior version first |
| No insight into search quality | Trial and error required |
| Data gravity | Even if replicated, user memories don’t transfer |
Strategic Position
The Gap We Fill
| Platform | Their Memory | The Limitation |
|---|
| OpenAI | ChatGPT memory | Only works in ChatGPT |
| Anthropic | Claude memory | Only works in Claude |
| Google | Gemini memory | Only works in Gemini |
| Microsoft | Copilot memory | Only works in Microsoft ecosystem |
| aiConnected | Brain | Works everywhere |
Brain Within aiConnected
Brain is not a standalone business. Brain is the foundation of the cognitive operating system.
aiConnected Cognitive Operating System
│
├── Brain (memory, continuity, learning)
│
├── Knowledge (retrieval, expertise)
│
├── Voice (verbal communication)
│
├── Chat (text communication)
│
├── Personas (identity, behavior)
│
└── Future: Vision, Motor Control (robotics)
Long-Term Vision
aiConnected is building the cognitive operating system for the physical AI era.
Today’s products (Brain, Knowledge, Voice, Chat) generate revenue and establish the memory layer. Tomorrow, this same infrastructure powers robotics:
| Component | Role in Robotics |
|---|
| Brain | Remembers tasks, learns from experience, maintains continuity |
| Knowledge | Accesses manuals, procedures, domain expertise |
| Voice | Verbal interaction with humans |
| Chat | Text-based commands, logging, reporting |
| Persona | Consistent personality, appropriate behavior per context |
Revenue Comparison
| Business | Revenue at Scale |
|---|
| Agency platform (100 agencies × $500/mo) | $600,000/year |
| Brain (100,000 users) | $29,520,000/year |
Revenue Layers
| Layer | Role |
|---|
| Brain | Foundation, entry point, highest retention |
| Personas | Monetization layer on top of Brain |
| Agency Platform | Cash flow engine, funds Brain development |
| Knowledge/Voice/Chat APIs | Developer and platform revenue |
| Robotics Cognitive Layer | Future enterprise and manufacturer revenue |
The Moat
-
Data Gravity: User memories accumulate in Brain. Switching means losing them.
-
Persona Relationships: A Writing Partner with 8 months of context on your book can’t be replicated elsewhere.
-
Cross-Platform Freedom: Ironically, Brain lets users use any LLM they want, but Brain itself becomes irreplaceable.
-
Network Effects: Teams tier means shared memories. A company’s institutional knowledge lives in Brain.
-
Ecosystem: Agencies building on aiConnected infrastructure aren’t switching because OpenAI added memory to ChatGPT.
Summary
Brain is the persistent memory layer for all AI. It solves the fragmentation problem that major platforms will never solve because doing so conflicts with their business model.
At scale, Brain alone generates $290M+ annual profit. Combined with Personas and the broader aiConnected product suite, the business supports the long-term vision of becoming the cognitive operating system for AI and robotics.
The protection strategy is trade secret over patent. Secrecy buys time; data gravity makes the moat permanent. By the time anyone reverse-engineers a comparable system, users have years of memories they won’t abandon.
aiConnected: All AIs connected through persistent memory.
Document Version: 1.0
Last Updated: April 2026
