About Maro Labs

A Korea-based software studio, founded in April 2026, making AI-native workspace tools for macOS.

Maro Labs was founded in South Korea in April 2026 by Maro. Building software with AI every day, we kept running into the same problem: the tools lost context along the way, and the decisions behind the work disappeared into old chats.

Maro Portal began as the workspace we needed for our own work: one place where tasks, conversations, and decisions stay on the record and nothing closes without review. A question in May about how to start a martial-arts novel grew into its writing extension, and in September 2026 into a separate app, Maro Writer.

We believe everyone who does serious work with AI needs the same thing: context that stays, work that is checked, and decisions that can be traced.

The problem we work on

  • Context gets lost

    Each AI session starts fresh. Plans, constraints, and earlier decisions have to be explained again, or they quietly get dropped.

  • Work is scattered

    Results live across many chats, tools, and folders. It is hard to see what was done and why.

  • AI work is hard to check

    When several AI tools work on one project, it is hard to review what each produced and keep them moving in one direction.

Record, review, decide

Every piece of our work runs through the same six-step loop inside Maro Portal. It is recorded, checked by a separate reviewer, and closed only with a decision on the record. Maro leads and decides; AI workers plan, build, and review in parallel.

  • 1

    Born

    A task is opened from the conversation, with your request kept in its original wording.

  • 2

    Checked

    Before it runs, the task is checked, so work starts from a clear request rather than a guess.

  • 3

    Run

    An AI worker does the work in its own session.

  • 4

    Evaluated

    A separate reviewer checks the result. The review receipt names the provider, and a partial pass stays partial.

  • 5

    Harvested

    Outputs are filed in dated work folders, and lessons go to the work encyclopedia.

  • 6

    Closed

    The task closes with its record. Nothing counts as done until it has been through review.

On the record (October 8, 2026)

closed work records
470
closed after review
424
test files
493
product build receipts
117

How we build

  • Evidence first

    Look at the real thing, then reason, then conclude. A claim of "done" is checked again, not taken on trust.

  • Done means usable

    If a feature works but is hard to use every day, it is not finished. Done is usable, correct, safe, and verifiable together.

  • Use what exists

    Before building something new, find what is already there. Wiring comes before new parts.

  • Calm above, busy below

    The first screen stays quiet. The complicated work happens under the surface, and you go to it only when you need to decide something.

  • Rules guide, bans are explicit

    Rules give direction rather than lock things down. What must never happen is written out plainly.

  • Failures are assets

    A failure is recorded, not hidden: what was expected, what happened, and the smallest check that would have caught it.

Built with Claude

Claude is our main tool. Over 70% of our important work, meaning direction, design, planning, and the key reviews, runs on Claude. Other AI tools take the rest, mainly implementation and second reviews.

Claude is also connected inside both apps: as a provider in Maro Portal, and for co-writing in Maro Writer. Maro Writer was specified in a Claude Code session on September 27, 2026.

Model-neutral by design

Maro Portal treats AI providers as replaceable parts, so other models also implement and give second reviews.

Company facts

Company
Maro Labs
Founded
April 2026
Founder
Maro
Based in
South Korea
Team
Founder-led, working with parallel AI workers
What we make
AI-native workspace tools for macOS
Languages
Korean, English (website also in Japanese)

Want to use Maro Portal or Maro Writer?

Tell us which app and what for. We reply by email.