Live Document

AI Should Get Smarter
With You, Not About You

These aren't marketing promises. They're architectural decisions enforced by code, verified by audits, and visible in our open-source repositories.

The Year is 2026.
Your AI Still Forgets Everything.

Every conversation starts from zero. Every coding assistant makes the same mistakes. Developers call it "context rot" — the frustrating reality that AI tools are stateless by design.

Research from METR found that traditional AI coding assistants create a 19% productivity loss because small context windows force constant manual prompting that destroys flow state.

We built ekkOS because memory shouldn't be a luxury — it should be infrastructure. And infrastructure should respect the people who depend on it.

60,000×
AI intelligence has scaled since ChatGPT launched
100×
Memory capacity has scaled in the same period
600×
The memory gap that ekkOS exists to close
the_problem.ts
// Traditional AI: stateless by design
session_1: "I prefer TypeScript strict mode"
session_2: "Our API uses snake_case"
session_3: "We fixed that auth bug yesterday"
// Session 4: AI has forgotten all of it
// ekkOS: memory as infrastructure
ekkOS.remembers("all of it")
ekkOS.learns("from mistakes")
ekkOS.getsSmarter("over time")

"Context Engineering" Is Now an Entire Industry

Developers spend hours structuring prompts, manually segmenting code into chunks, and fighting the "lost-in-the-middle" problem where AI loses track of details buried in conversations. This shouldn't be your job.

🔄

The Repetition Tax

Every session starts over. You re-explain your codebase, your preferences, your conventions. The AI that helped you yesterday is a stranger today.

📉

The Paradox

More context should help. Instead, output quality degrades as you add more — models pull irrelevant details from earlier prompts and lose focus.

💸

The Hidden Cost

A KV cache miss costs 10× more than a hit. Agentic AI generates 100× more tokens. Without smart memory, inference costs explode.

Four Principles. Zero Compromise.

1

We Cannot See Your Memory

Not "we won't" — we can't. Row-level security is enforced at the Postgres layer. Every query is physically incapable of returning another user's data. Your memories are encrypted with keys derived from your identity. Even our engineers with full database access see only encrypted blobs they cannot decrypt.

Verified by production security audit
2

Your Memory Travels With You

Full export. Full portability. One click downloads everything — patterns, episodes, directives, secrets (encrypted) — in documented JSON format. Switch to a competitor tomorrow and take every byte with you. We'd rather earn your loyalty than trap you into it.

ekkOS_Export available in every tier
3

Collective Learning, Individual Privacy

When you discover something valuable — a debugging pattern, an architectural insight, a gotcha — you can share the structure without sharing your data. "Always check for null before accessing nested properties" helps everyone. Your actual code stays yours.

Automatic PII scrubbing on collective patterns
4

Delete Means Delete

When you delete something, it's gone. Not archived. Not soft-deleted for 90 days. Not retained for "business analytics." Purged from primary storage immediately, from backups within 30 days. Cryptographically unrecoverable. Your right to be forgotten is a button click.

GDPR Article 17 compliant by default

The "Never" List

Some things aren't negotiable. Not for investors. Not for acquirers. Not for anyone.

💰

Never: Sell Your Data

Not to advertisers. Not to data brokers. Not to AI training companies. Not for any amount of money.

🧠

Never: Train on Your Memory

Your memories don't train our models or anyone else's. Your intellectual property remains yours.

👤

Never: Build Shadow Profiles

No tracking non-users. No data collection from people who haven't explicitly signed up.

🪤

Never: Use Dark Patterns

No tricks to keep you subscribed. No buried cancel buttons. No "are you sure?" guilt trips. Cancel in two clicks.

📈

Never: Surprise Price Hikes

Your tier is your tier. Changes announced 90 days minimum. Grandfathered pricing for existing customers.

🎯

Never: Weaponize Memory

No selling to surveillance companies. No government backdoors. Memory is for helping people, not hunting them.

How We Actually Make Money

When a company's business model is unclear, you're usually the product. Here's exactly how ekkOS generates revenue.

$

Subscriptions

You pay for compute, storage, and features. Free tier for individuals. Pro tier for power users. Team tier for organizations. The value exchange is clear: you pay money, we provide infrastructure.

B

Enterprise Contracts

Organizations deploy ekkOS for their engineering teams. Custom SLAs, dedicated support, SSO integration, on-premise deployment options. Volume pricing for large deployments.

That's It

No advertising revenue. No data brokerage deals. No "strategic partnerships" that compromise your privacy. No investor pressure to monetize your data. Just software that works, sold honestly.

Policies Enforced by Architecture

Trust shouldn't require faith. Every commitment below is enforced at the infrastructure level — not by policy documents, but by code you can audit.

CommitmentTechnical EnforcementStatus
Complete data isolationPostgreSQL RLS on every table. Queries physically cannot return other users' data.Verified
Encryption at restAES-256-GCM with per-user key derivation. Secrets layer has additional envelope encryption.Verified
Immutable audit trailEvery write creates evidence chain entry. Append-only. Tamper-evident logging.Verified
Rate limitingPer-tier limits at API gateway. No exceptions, including for internal requests.Verified
Hard deletionDELETE CASCADE on foreign keys. Backup rotation ensures complete purge within 30 days.Verified
Built on Open Standards

MCP: The Memory Protocol

ekkOS implements the Model Context Protocol (MCP) — an open standard for AI memory infrastructure. This means you're not locked into our implementation. Any MCP-compatible system can read your exported data.

We believe memory infrastructure should be as interoperable as HTTP. Proprietary lock-in is how companies extract value from users instead of providing it.

mcp://
ekkOS_Search
ekkOS_Forge
ekkOS_Recall
ekkOS_Export
+ 27 more tools

31 MCP tools, all documented, all auditable

Open by Default

Our VS Code extension, SDK, and MCP implementation are open source. Read every line of code that touches your data. The protocol we use is documented and standardized. No black boxes. No "trust us" security.

Questions? Skeptical? Good.

We built ekkOS for people who read the fine print. If anything here is unclear, or you want to verify a claim yourself, we're here to help.

Last updated: January 2026 · Version 2.0 · View commit history