Agentryx AIDWG 00 · INDEX
Operator-led · Agent-first · 2026AGENTRYX-AI.COM · EN / 한국어

Four products.
Built and run by agents.
An operator decides what ships.

Agentryx AI is an operator-led AI studio. Agents handle the building, reviewing, shipping, and operating — end to end. One operator sets priorities and decides what ships.

See the productsRead the operating thesis →
Sheet index01ProductsM-01–0402ThesisI–IV03Why nowOBS 01–0304Contact→
01

Four bets, one operating system.

Each product is a thesis strong enough to ship.
M-01 · AGENTRYXIn operation

Agentryx

The tool this studio runs on

Turns a goal into finished work on its own — the same recursive rule at every level, replans instead of retrying blindly, and never calls something done until the evidence proves it.

PatternRecursive orchestrator-workers
RunnerClaude · Codex CLI
IsolationPer-task git worktree
CoordinationSingle-conversation coordinator
The task model — run-or-split, run, synthesize — runs the studio's real work daily, and goal-achievement verification decides what counts as done.Product file →
M-04 · RETALKLaunched

ReTalk 리톡

The conversations KakaoTalk lets slip away — restored, searched, kept. Local-first; nothing leaves your device.

StanceLocal-first by design
PlatformWindows 10/11 · macOS 13+
License₩49,000 · one-time
retalk.agentryx-ai.com ↗Product file →
M-02 · ITINEVAIn development

Itineva

A travel planner that checks its own work — every place verified as open and real, every day's schedule checked against actual travel time — before it hands you a shareable trip page.

OutputVersioned trip page
SharingOne link, always current
MVPComplete · not yet released
Product file →
M-03 · MODUBOZAIn development

ModuBoza 모두보자

Find something to watch and pay less for it — streaming discovery with one simple checkout. Clear price up front; leaving is as easy as joining.

FocusStreaming discovery
CheckoutOne simple payment
MVPComplete · not yet released
Product file →
02

How a small studio runs many products.

Read the operating thesis →
01

An operator should run the company.

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

02

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.

03

We'd rather verify than rush.

We build static checks and human judgment into how work ships, and we don't treat a green check as finished.

04

Ship the smallest verifiable thing.

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

03

Three observations we are building against.

Observation 01

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

We're testing 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.

Generic tools force every studio into the same workflow. One built in-house can be shaped exactly to how the work gets done.

See for yourself

Judge us by what ships.

Try each product as it launches, and hold us to the thesis. If you want to compare notes on running a company with agents — the operator reads every email.

Get in touch