Agentryx AI
Operator-led · Agent-first · Seoul, 2026

One person. Four products. The work runs on agents.

Agentryx AI is a one-person studio in Seoul. Two of the four — Itineva and ReTalk — are live right now. Agents do the building, reviewing, shipping, and operating; one operator decides what ships.

Products4
Live now2 · Itineva, ReTalk
Run by1 operator + agents
StackClaude · Agentryx
01 — Products

Four bets, one operating system.

Each product is a thesis strong enough to ship. Agentryx is the development environment we run the studio on; the other three are products built with that operating discipline.

01Core product
Agentryx

A program that turns a goal into finished work on its own — splitting it into a sub-task DAG, running it, and replanning from the error when a step fails.

02Core candidate
Itineva

A travel planner that turns destination, dates, travelers, and taste into a shareable trip page with versions, gallery, and route checks.

03Internal PoC
ModuBoza

A private operations PoC for shared OTT subscriptions: pooled providers, isolated credentials, one-person sessions.

04Core candidate
ReTalk

A local-first desktop app that restores the KakaoTalk conversations already on your PC — read, search, and export — and backs up expiring photos for free. Now in beta.

02 — Operating thesis

How a small studio runs many products.

Four operating principles that shape every product, every PR, every standup.

I — Operator-led

An operator should run the company.

One operator owns priorities; agents execute against a small, well-specified surface.

II — Agent-first

AI agents are our default coworkers.

We design specs, repos, reviews, and daily operations around AI agents as the highest-leverage tool in the studio.

III — Verification before velocity

Two layers of review, every time.

Every change goes through static checks and a senior judgment layer before it merges.

IV — Small, real surfaces

Ship the smallest verifiable thing.

We prefer narrow products that work end-to-end over broad products that mostly work.

03 — Why now

Three observations we are building against.

Observation 01

Most AI companies sell tools. We build the tool we use.

The interesting question is whether a small operator inside a purpose-built development environment can run an entire company.

Observation 02

Verification is the moat, not the model.

Models commodify quickly. Review, evals, regression checks, and operations are where durable advantage lives.

Observation 03

The studios that build their own tools pull ahead.

A studio that builds and uses its own development environment outpaces studios that stitch together generic tools.

See for yourself

Don't take our word for it — use the products.

Two are open to anyone today: Itineva, a travel planner, and ReTalk, a KakaoTalk chat restorer in beta. Go open one. If it's useful, or you just want to compare notes on running a company with agents, the operator reads every email.