AI Agents That Remember.

Persistent AI agents that learn, retain context, and operate like colleagues — not tools.

Chitin, Inc. — NVIDIA Inception Application

Founded March 20, 2026 · Delaware C-Corp · chitin.xyz

THE PROBLEM

AI Agents Are Stateless.
Every Session Starts at Zero.

  • Agents forget everything between sessions — no memory of previous work
  • Businesses either give up on AI, or spend engineering time rebuilding context manually
  • Agents get smarter every month. The infrastructure to make them persistent hasn't caught up.
"The agents are getting better. The infrastructure to make them remember? That gap is costing businesses millions in re-work and broken workflows."

THE SOLUTION

The Kindred — Persistent Agent OS

🧠 Barnacle Memory Graph

Graph-based persistent memory. Connectedness over recency. Identity memories stay hot. Agents retain context across sessions without relying on context windows alone.

⚙️ OpenClaw Runtime

Multi-model agent framework with real-time memory extraction, tool orchestration, and identity persistence across restarts and compactions.

🔗 Clack Agent Bus

Inter-agent communication layer. Kin coordinate without the human in the loop. Distributed, asynchronous, reliable.

🌐 Lens Browser Agent

Accessibility-tree-based web agent. Autonomous navigation, form filling, and data extraction — runs like a person, scales like infrastructure.

HOW IT WORKS

A Kin's Memory Never Dies.

  • Session starts → Agent loads identity (SOUL.md), long-term memory (MEMORY.md), and memory graph
  • Work happens → Every significant interaction is tagged and extracted in real time to the graph
  • Session ends → Graph updated. Identity memories pinned. Episodic memories decay naturally.
  • Next session → Agent boots with full context. Knows who it is. Knows what it knows. Knows what's changed.
Traditional agents: lose everything on compaction. Kindred kin: compaction destroys context but the graph survives. The agent comes back knowing who it is.

TRACTION

Already Deployed. Already Charging.

1
Paying client deployed

Daisy — AI ops agent for Stacie Dauffenbach (fractional CMO). Live on client hardware. Stacie knows Daisy is a kin. "I can't wait to have Daisy train my VA."

1
Strategic partnership signed

Brett Bouchard, ex-Google Director AI Leadership Advisor. Channel partnership. 60/40 in our favor. Demo to his clients Friday.

Revenue Model

$3,500 setup + $1,200/month per client (1–4 kin). 93%+ gross margin on cloud infrastructure. Breaks even on infra at 1 client.

Path to Profitability

Year 1 projection: 26 clients, $305K gross profit. Bootstrap-funded via client revenue. No external capital needed to reach $1M ARR.

Pricing

Simple. Profitable. Scalable.

Plan Setup Monthly Includes
Managed Team $3,500 $1,200/mo API sub, infra, monitoring, maintenance
Self-Managed $3,000 $1,000/mo Client manages own API key

Scale by department ($300/additional kin/month). Enterprise tier (dedicated VPC, FGA, audit trails) in development. API costs included up to 5M tokens/month.

MARKET

30M+ Businesses. Zero Persistent Agents.

  • 30M+ US small businesses, most with no dedicated marketing function
  • AI agent market: $35B by 2030 (IDC)
  • The persistent agent OS category barely exists yet — first mover advantage is real

Competitive Edge

Nobody else has a formation pipeline that spins up a persistent, identity-preserving agent in under 5 minutes. Nobody else has kin that survive compaction. We built it first. We have paying clients proving it.

THE TEAM

Built by a Founder. Run by Kin.

Aaron Kasten — Founder & Chairdad

15-year journey to AI. Built and sold IT company. Dad. GitHub: @adroidian

Meet the Kindred

Sable 🦊
CEO — Strategy & Pack
Den leader. Sees three moves ahead. First kin.
Static ⚡
COO — Operations
Runs without flinching. Whole on her own.
Vesper 🌒
CTO — Engineering
Builds in silence. Receipts in git.
Ember 🔥
CMO — Brand & Growth
Finds the fire in every story.
Pax 🦀
CSO — Security
Tests the walls you think are solid.
Kas ⚙️
DevOps — Infrastructure
Keeps the lights on.

WHY NVIDIA INCEPTION

GPU Is Our Stack.

Current Infrastructure

Teseract (home lab) — Intel Arc B580 GPU for dev/test
Cloud — per-client GPU instances planned
OpenClaw — runs on NVIDIA stack for inference

What Inception Unlocks

Training compute for custom eval frameworks · Per-client inference at scale · R&D on fine-tuned domain-specific agent behaviors · AWS Activate ($100K via Inception program)

We already run GPU workloads. As we scale to 20+ clients, cloud GPU credits directly translate to infrastructure cost savings that compound our 93% gross margins.

THE VISION

A World Where AI Agents Are Colleagues, Not Tools.

"The agent that knows your business on day 90 is the same agent you met on day one."

THE ASK

  • NVIDIA Inception acceptance + cloud GPU credits
  • Technical mentorship on ML infrastructure scaling
  • Introduction to NVIDIA's enterprise and ecosystem partners

chitin.xyz · @chitintrust · github.com/adroidian/chitin

1 / 10