MK Code
MK Code plans the work, writes the code and checks the result in a real browser, all on models you control.
Describe the task for the agent…
This is the session from the top of the page, as it happened in the app.
01Plan
In plan mode the agent reads the project and lays out the steps without touching a file. You decide when it starts.
Describe the task for the agent…
02Build
Implement the plan and the agent gets to work. You can follow every file it reads and edits as it goes.
Describe the task for the agent…
03Check
The agent opens the page in the browser built into the app, clicks through it and fixes what looks wrong.
Agent is opening the page
Opened http://localhost:8123/
04Commit
Pick the files, keep the commit message it drafts from the diff or write your own, and commit to the branch.
Open up to four panes and give each one its own task. With worktree on, an agent works on its own branch, so parallel changes never collide.
Ask how something works and the agent sketches it: flow diagrams, notes and file references on a board you share.
Save a prompt and run it on a schedule while the app is open.
Name
Weekly ridership summary
Next runs
Ask questions, draft emails and work through ideas with the same models, outside any project.
Message MK Code…
Bring in GitHub, Linear, Notion, Sentry, Figma and more through MCP, or add your own server.
MK Code runs on the models you deploy through MeetKai, in your own environment. People sign in through the browser with their account, and the credential stays on their computer.
A narrated walk through MK Code, from signing in to agents working in parallel.
Request a demo