AgentBase — Seoul

This company runs on agents.

We do not hire people for the operating roles. We build the agents that hold them, and we run our products on that fleet.

22
agents in the fleet
2
people
7
approval gates

Each square is a unit of operating work. Move the cursor: scattered work falls into formation. Colour mix mirrors the fleet — 16 domain, 3 meta, 3 watchdog.

01 — The fleet

Twenty-two agents, named and accounted for.

Not a free-roaming loop. Each one is a function a durable workflow invokes for a judgment-heavy sub-task: a system prompt, a curated tool subset, an output contract, a spend cap, and a defined escalation.

Domain16

Do the operating work.

  • sourcingBrief to ranked candidates
  • vettingFit score and brand-safety flags
  • outreach_writerGrounded personalised outreach
  • conversationClassify and answer replies
  • conversation_responderDraft the reply that goes out
  • logisticsCreate and track shipments
  • content_verifyFind the post, match the brief
  • analystCompile and narrate the report
  • researchSite to sales analysis
  • intakeOne line to a valid brief
  • lead_outreach_writerOutbound to brands
  • payment_mandateAuthorise spend against a mandate
  • complianceDisclosure and policy checks
  • creativeBrief-aligned creative direction
  • a11yAccessible copy and structure
  • customer_successActivation and retention

Meta3

Route, score, and tune the ones that do.

  • coordinatorRoute work across the fleet
  • criticScore before anything ships
  • optimizerTune what the critic scored

Watchdog3

Watch drift, spend, and the trust boundary.

  • anomaly_watchFlag runs that drift
  • cost_watchStop spend before the cap
  • security_watchWatch the trust boundary

02 — The replacement ledger

Eight seats a person used to sit in.

The left column is quoted verbatim from our internal agent roster, written before this website existed. We did not rewrite them to read better. They read like someone's actual job because they were — ours.

  1. before · humanthe human clicking Search + Step 2
    after · agentsourcingBrief in, ranked candidates out, de-duped against the blacklist and every prior campaign.
  2. before · humanthe human eyeballing Step 2's table
    after · agentvettingPulls each profile, computes engagement and average views, scores fit, raises brand-safety flags.
  3. before · human"AI 작성" button + the human reviewing
    after · agentoutreach_writerWrites multiple angles, runs a four-judge tournament, checks spam score, returns one draft.
  4. before · humanthe human reading the Replies tab
    after · agentconversationClassifies every reply, extracts the address or the rate, drafts the response or escalates.
  5. before · humanthe human in Step 4
    after · agentlogisticsCreates the shipment, watches tracking, handles the exceptions.
  6. before · humanthe human in Step 5
    after · agentcontent_verifyPolls for the post, matches it against the campaign brief, computes what it did.
  7. before · humanthe human in Step 6
    after · agentanalystCompiles the report and writes the narrative around it.
  8. before · humanthe 6-tab campaign-create form
    after · agentintakeA short conversation that turns a one-line ask into a valid campaign brief.

…and 14 more agents doing work that had no human predecessor at all.

Quoted from our agent roster.

03 — How the work moves

Autonomy is a setting, not a personality.

Work moves through named stages, and every irreversible action sits behind a named gate. Gates ship on by default. An operator moves the whole workspace between three levels — and the level, not the mood of a model, decides what happens without a human.

Stages

  1. 01overviewBrief accepted, plan generated
  2. 02sourcingFind and vet creators
  3. 03outreachWrite, send, handle replies
  4. 04shippingShip samples
  5. 05content_reviewVerify posts went live
  6. 06performanceCompile the report

Gates

  • approveShortlistafter sourcing + vetting
  • approveOutreachSendbefore sending each batch
  • approveReplyResponsebefore sending a drafted reply
  • approveShipmentrequiredbefore creating a shipment
  • approveStageAdvancemoving the campaign to the next stage
  • approveContentrequiredfinal sign-off on a verified post
  • approveBudgetrequiredreleasing spend, executing a contract

Autonomy levels

  • copilotagents only propose; a human clicks every action
  • checkpointeddefaultagents act, each stage has an approval gate
  • autonomousagents act end to end; escalate on policy exceptions

04 — The operating layer

$2,400 and 45 hours became $7.40 and 2 hours.

The same piece of work, priced two ways. The left column is what an outside operator charges to run it, plus the hours it takes them. Ours is metered agent and infrastructure cost, plus the hours a person spends at the gates.

The management and operations layer only. The spend that passes through to third parties is identical either way and is excluded — including it would make the difference look larger than it is.

Outsourced to an operator

$2,400

operating cost

20% management fee on $12,000 of pass-through spend

45

human hours

Run by the fleet

$7.40

operating cost

metered agent and infrastructure cost

2

human hours

$12,000 — Pass-through spend — identical on both sides · excluded from this comparison

One engagement of the size a small team would outsource.

Have a workload shaped like this? Send one line →

05 — How we staff

We don't hire people. We build agents.

That is a statement about where headcount goes, not about who is accountable. Two people run this company, and the boundary between what they decide and what the fleet executes is written down, versioned, and enforced in code.

What stays with a person

  • Anything irreversible: releasing spend, signing a contract, final sign-off on published work.
  • Setting the autonomy level, and moving it.
  • Deciding what the company builds next.
  • Everything on this page. Two names, at the bottom.

Gates a person can never delegate: approveShipmentapproveContentapproveBudget

What the fleet holds

  • The operating roles a company this size would otherwise hire for.
  • The work that runs on a schedule, and the work nobody wants to do twice.
  • The checks on its own output — scoring, cost, drift, and the trust boundary.

06 — The company itself

The same rule applies inward.

Merging to main is the release — no human runs a deploy command. Deploys to the agent service land as a zero-traffic canary and promoting one stays a deliberate human decision: the same gate-shaped governance the products are built on.

Gates on every pull request5

  1. verify-build (lint, build, type-check)
  2. TypeScript unit suite
  3. offline Python suite + golden-eval holdout gate
  4. submission artifact suite
  5. checksum-pinned secret scan

Test suites that block the merge

3,000
Python
596
TypeScript

This website is in that loop too. The figures on this page are a JSON snapshot committed to a public repository and rendered on the server — nothing here is fetched after the page loads.

07 — What we operate

The fleet is not the product. It is how the products get run.

Each one makes its own case on its own site.

08 — The lab

What else two people shipped.

The fleet frees the calendar, and the calendar fills with builds. Everything below is verifiable without taking our word for it — a live site, a public listing, a demo video, or a judged result.

Also on the bench

  • Fairthonthe six-hat evaluation system Glasshat grew out of, scoring pitch decks and repositories against one rubric
  • VibeMeetinga macOS overlay that translates the other side of an English call and drafts three answers you can say yourself
  • ClaudeSynckeeps AI coding environments in sync across Macs
  • vibeVoicethe text-to-audio dashboard our demo narrations are made with
  • openClawWorlda 2D world where people and agents share the map
  • Memoed on your lifean evidence-first iPhone app that finds what changed in everyday memories

09 — The competition record

The whole record, results as they fell.

Public competitions are the cheapest neutral benchmark a two-person company can buy: outside judges, fixed deadlines, published winner lists. We enter with the fleet and publish the whole column — the ribbons and the losses alike.

12
entries
1
1st place
1
Honorable Mention
1
Judging
  1. Gemini 3 HackathonGoogle
    Somm.dev
    No award
    Devpost
  2. PlayMCP “Player 10”Kakao
    kidsafe-mcp
    Selected

    Program selection. No public roster page exists, so this row rests on our own records.

  3. Gradient AI HackathonDigitalOcean
    vibeDeploy
    1st place$8,000
    winnersDevpostvideo
  4. Gemini Live Agent ChallengeGoogle Cloud
    VibeCat
    Honorable Mention$2,000
    winnersDevpostvideo
  5. GitLab AI HackathonGitLab
    GitLab Atlas
    No award
    Devpost
  6. Built with Opus 4.7Cerebral Valley × Anthropic
    Preview Forge
    No award
    videosite
  7. Mod Tools HackathonReddit
    vibe-mod
    No award
    Devpostlisting

    No award; the app itself passed review and is live in the Reddit App Directory.

  8. Vector Space DayQdrant
    Memex
    No award
    site
  9. Rapid Agent Hackathon · Arize trackGoogle Cloud
    Glasshat
    No award
    Devpostvideo
  10. AI Agents Challenge · Track 3Google for Startups
    SocialSeed.ing
    No award
    video
  11. Games with a HookReddit
    Pipuzzle
    No award
    Devpostlisting
  12. Global AI Hackathon Series with Qwen CloudAlibaba
    Circle Take
    Judging
    Devpost

The last row is still being judged — the result lands on Aug 21, 2026, and this table will carry it either way.

Judged once: CMUX × AIM Intelligence Hackathon, Seoul, April 2026 — our own record, not a published roster.

And two internal hackathons of our own, scored against the same kind of rubric.

10 — What we can hold for you

Three ways to put the fleet on your work.

Not a promise of outcomes — a choice of engagement shapes. Each one points at the part of this page that backs it.

Send one line →

11 — Where we operate

Operating work does not need a local office.

12 — Intake

One line is enough.

The fleet has an agent named intake, and its whole job is turning one line into a valid brief. Send a line about your work — the thing that runs on a schedule, the thing nobody wants to do twice. Both founders read every line.

No form? The addresses in the footer work the same.

Grants & programs

ElevenLabs Grants

Built with

  • Google Cloud
  • Google Gemini
  • OpenAI
  • Vercel
  • Next.js
  • MongoDB
  • Go
  • Tailwind CSS