KRContact

Korean-first AI coding agent

Mirr Code

A Korean-first AI coding agent

A coding agent you talk to in the terminal, in Korean. It runs the tools and stays with the task until it is done.

Our own model

We build the model ourselves.

Mirr Code runs on models that Patty trains and serves itself. There is no external model API behind it, so Korean is not a translation target but the standard for training and evaluation.

Cursor movement during Hangul composition, choseong search, reading mixed-language repositories, and understanding requirements in public documents all happen with no separate translation step.

See the Mirr model
Request to completion

One request, finished in a single session.

read business-plan.hwpx, fill in the missing permission checks, and run the tests
  1. 01
    Understand the Korean requirements

    Read body text, tables, and outlines in HWP and HWPX documents to determine what needs to be done.

  2. 02
    Review the repository and team rules

    Review project structure, existing code, notation, and team guidance to decide what to change and how.

  3. 03
    Plan and confirm authority

    Propose which files change and why first; work that requires approval is confirmed before it runs.

  4. 04
    Change and verify the code

    Edit the code and run type checks and tests in the same session. When a problem arises, it finds the cause and fixes it.

  5. 05
    Summarize the result

    Changed files, the checks that ran, and anything that still needs review are laid out for you.

Korean in the product

Mirr is designed for the Korean working environment.

From input and search to document understanding and result explanations, the whole development workflow in Korean is taken into account.

01

Natural Hangul input

Composing Hangul is never mistaken for committed input. Frequent actions during Hangul input, such as cursor movement and deletion, are continuously tested.

02

Fast search by choseong

Commands can be found in Korean, English, or by initial consonants alone. When several commands match, the options are shown instead of running one arbitrarily.

03

HWP and HWPX documents, understood right away

Body text, tables, and outlines in HWP and HWPX documents are read to find what the work needs. Encrypted documents are never guessed at; it reports that they cannot be read.

04

Korean expressions in the repository, kept as-is

In repositories where Korean and English are used together, it finds the terms and notation already in use and follows them as they are.

Runs to the end

How Mirr works

Check the information needed, plan the change, then execute and verify within defined permissions.

  1. 01

    Check what the work requires

    Review code, configuration, documents, repository guidance, and the current work state to identify what is needed.

    repository · documents · instructions
  2. 02

    Plan the change

    Break the goal into steps and show which files will change and how they will be verified before execution.

    plan · impact · validation
  3. 03

    Execute within defined permissions

    File edits and tool execution happen only inside the allowed permissions and sandbox. Work that needs additional authority requests approval first.

    permissions · sandbox · approval
  4. 04

    Revert when problems arise

    When something goes wrong mid-task, it returns to the previous state and proceeds again, checking test results to fix the cause.

    checkpoint · rewind · recovery
  5. 05

    Leave results you can check

    What changed, the checks that ran, the outcome, and what still needs review are kept as a record.

    diff · test · DARI receipt
Execution modes

Shift+Tab decides how much you hand over.

The mode can change mid-session and applies from the next tool call.

01

Plan

Read-only. It changes nothing and only works out what it would do. You read the plan, then switch modes.

02

Normal

Every tool call is approved as it happens. It asks before each file edit and each shell command.

03

Auto

Writes are delegated. The work runs inside the sandbox and stays there. Anything that has to cross that boundary still comes back for approval.

04

YOLO (Ctrl+Y)

It runs to the end without asking. Use it for long repetitive work you do not want to sit through. Questions and plan approvals still reach you, and the checkpoints remain so it can be rewound.

switch with Shift+Tabreadwriteshell
planRead-only. Produces a plan, changes nothing.yesblockedtrusted only
normalEach tool call is approved as it happens.yesapproveapprove
autoWrites are delegated, inside the sandbox.yesyesyes

YOLO (Ctrl+Y) still stops for questions and plan approvals

Commands

Commands answer to Korean names.

There are 41 slash commands. The Korean name, the English name, and the choseong initials all find the same one.

ㅅㅋpress / to open

KoreanEnglishChoseong
/스킬설정/skillsㅅㅋㅅㅈ
/샌드박스/sandboxㅅㄷㅂㅅ
/사용량/usageㅅㅇㄹ

41 commands · an ambiguous entry lists candidates instead of running

Rewind
/돌아가기 · /rewind

※ The initials ㄷㅇㄱㄱ find it too.

Usage
/사용량 · /usage
Skills
/스킬설정 · /skills
Korean compliance
/컴플라이언스 · /pipa · /kisa · /csap
Sandbox
/샌드박스 · /sandbox
All commands
41

※ Press / to open the palette. An ambiguous entry lists candidates instead of running one.

Built for Korea

Elsewhere these mean a converter, or giving up.

The things Korean teams need, built in.

01

HWP and HWPX documents

Body text, tables, and outlines are read straight into the working context. No separate converter. An encrypted file is reported as unreadable rather than guessed at.

02

PIPA, KISA, CSAP checks

One /컴플라이언스 command runs the review from each of those angles. /pipa, /kisa and /csap reach the same place.

03

Choseong search

Commands and files are found by Korean initials. ㅁㄷ is enough to reach the model commands.

04

Hangul input itself

A syllable still being composed is not mistaken for a finished one. Cursor movement and deletion have their own path and regression tests.

Technical review

How Mirr is built

See how Mirr is structured and the environments it runs in.

One runtime, many ways to run

CLI, TUI, headless, desktop, ACP, and HTTP servers all share the same principles. However you run it, tool calls, permission management, and checkpoints apply the same way.

Our own model and extensions

The model is developed and operated by Patty. Plugins, skills, MCP, agents, and ACP connections are available in the same runtime.

Why permissions and recovery are built in

Permissions are clearly managed so the agent works only within the scope it needs. Pre-execution confirmation, sandboxing, checkpoints, and rollback are part of the default workflow.

Plans

Two personal plans, by how you work.

Prices are read live from the checkout catalog.

Pro

The standard plan for steady daily use

₩49,000₩29,000/ month

  • Weekly and 5-hour usage allowances
  • 2 parallel sessions
  • HWP and HWPX document handling included
  • Every runtime (CLI, desktop, headless)
Start with Pro

Max

For long, unlimited sessions

₩99,000₩49,000/ month

  • No usage limits
  • 3 parallel sessions
  • Long-running sessions resume from checkpoints
  • HWP and HWPX document handling included
Start with Max

Start using Mirr Code.

Ask in Korean from your terminal, and the work begins.

Get started with Mirr Code