Early access · Windows
Maxtkn is a desktop workspace where agent teams plan, write, test and review real projects on your machine while you supervise, without reading code. They check their own work, and can keep going while you are away.
Maxtkn
Projects
Task Tracker
A task tracker where…
Add dark mode
Portfolio site
A task tracker where I can add, complete and delete tasks
Lead
I will break this into steps and brief the team.
Tester wants to run a command
node --check app.js
Task Tracker
2 items left
Plan the launch
Write the announcement
Book the demo
Works with the models you already use
The problem
If you describe software in plain language, you deserve to understand and steer what gets built, not approve diffs you cannot judge.
You get a wall of diffs and are asked to judge it. If you are not a programmer, you are approving blind.
Planning, building, testing and reviewing all happen in one long chat, so you cannot tell who did what, or why.
Changes land in your files with no clear owner and no easy way back, so every run feels like a risk.
Features
Everything an agent does is visible, attributed and reversible, so you can stay in charge without becoming a programmer.
A lead plans the work and hands tasks to specialists: architect, designer, builder, tester and reviewer. For bigger work the architect shows a design with diagrams first, and nothing is built until you approve it. Independent tasks run at the same time.
Independent tasks run at the same time.
Three access modes and four permission levels control what an agent may do alone. Risky commands, and secrets such as .env files, always stop and ask first.
Click a mode to try it.
Each file an agent writes shows who wrote it and exactly what changed. Keep or reject single hunks, revert one change, or rewind the whole project to before any message.
.card {
− padding: 8px;
+ padding: 24px;
+ border-radius: 16px;
}
A live preview, including your own project dev server, sits beside the conversation, next to the file tree, a code editor, the changes list and a real terminal. Draw on a picture of the page to show what to change, and when it looks right, publish the site as a zip or to Netlify after seeing exactly which files go.
Task Tracker
Plan the launch
Write the post
Book the demo
Each project is a folder. Each thread is one task. Run several in parallel, search every conversation, pin or archive them, fork one from any message, or ask a quick question with no project at all.
Each thread is one task. Run several at once.
Agents can only touch files inside the project folder you chose, and commands are checked before they run. You can also run them in a container or sandbox. Your keys sit in the Windows Credential Manager and your history stays on your device.
Documents/Maxtkn/task-tracker
agent ▸ write_file index.html
Written inside the project
agent ▸ write_file ../secrets.txt
Path must stay inside the project folder.
Checked on every file operation.
Checks
You should not have to read code to know if something works. Maxtkn runs the checks, looks at the running app, and shows you the evidence.
After the team edits, Maxtkn runs the project tests and build. A failure goes back to the team to fix, up to a limit you can see.
An agent can open your running app in a real browser, read the page, click, type and take screenshots. Addresses on your local network are blocked.
Write "the Add button adds a task" and get a stored check you can run again. Results show in a Checks tab.
The final walkthrough lists the evidence that was collected. It is built from what the checks found, not from the model grading its own work.
After a run a reviewer lists findings you can jump to. Fix one with a click, or dismiss it for good.
Ask for any change, error or command to be explained at the depth you choose. It never makes up details.
Evidence
ExampleWritten from the results of the checks. A check that could not run is listed as not run.
How it works
Watch a run from the first click to a finished app. Connect a model once and you can run the same flow yourself in minutes.
Illustration of the app. Hover to pause, or pick a step.
Agent teams
A lead plans and delegates. Specialists work in parallel. Duplicate any built-in agent, or use the team designer to pick agents, choose the lead and say who hands work to whom. Start from a template, share a team as a file, and see which of your teams have actually worked, with their success rate, usual time and cost.
1 agent
One Builder agent works directly. Fastest and cheapest for small changes.
3 agents
A Lead plans and delegates to a Builder, then a Reviewer checks the result.
6 agents
Lead, Architect, Designer, Builder, Tester and Reviewer. Specialists run in parallel.
Workspace
Point at what you want changed, manage your files, save versions and see what a run cost, without leaving the window.
In the Preview, click a button, heading or box, or drag a box around an area, then say what to change. The agent is told which element it is, where it lives in your files and which style rule paints it. Open the preview in its own window with phone, tablet, laptop and wide sizes.
Task Tracker
Pointing at the Add task button
Click an element, then say what to change.
Rename, move, duplicate, delete or create files and folders from a menu. Pick several with Ctrl or Shift and drag them onto a folder. Nothing is ever overwritten, and a deleted text file can be brought back from the Changes tab.
See what changed, choose what to include, save a version, switch branches, upload and download, and read the history, all in plain-language messages.
Add the task list and its styling
Every file lists who last changed it, when and why. Open a file to see its whole history with a link back to the conversation and to the saved copy from just before. It is built from what the team and you did, nothing is guessed.
Cards needed more breathing room
Open the conversation
Built from what the team and you did.
The message box says "About 40 cents · ~6 min", worked out from your own recent runs or marked as a guess. A finished run says what it took. Day and month budgets show what is left and when they renew.
Add a dark mode switch to the settings page
About 40 cents · ~6 min
The estimate comes from your own recent runs.
Save a thread as a record of who did what and when. Each step is linked to the one before it, so a change to the file can be found, and a script checks a record without the app.
Control
Choose an access mode per thread and a permission level per project. Whatever you pick, agents stay inside the project folder, and every change can be undone.
Every file change and every command waits for your approval.
When you want to see everything.
Agents create and edit files freely. Deleting files and running commands still need your approval.
The balanced default.
Files and commands run without asking. Dangerous commands still stop and ask.
When you trust the task and the team.
The access mode says when to ask. The permission level says what is possible at all.
Autopilot
Hand over a goal, a schedule or a pipeline, and come back to a report. Work that nobody watches runs under stricter rules than work you supervise.
Say "make the tests pass". The team works in rounds until a check that really ran says it is met, or a time, round or money limit stops it. Claims without a check are marked as claims, and each check says what it will look for before you start.
Hourly, daily, weekly or monthly tasks run while Maxtkn is open and report to an inbox. You allow each one explicitly, and that allowance goes stale if the task changes.
A thread can work in its own copy of a git project, on its own branch. Nothing reaches your project until you click Apply, and conflicts get a guided screen.
Run a whole request from a terminal or a CI job. Events print as JSON, and the exit code says whether it finished, failed a check or needed an approval it was not given.
When nobody is watching, the rules get stricter. A run on its own never waits for an approval and never grants one: anything that would ask is denied and listed in the report. It never uses Full access, and every run has a time limit and, for priced models, a money limit.
$ maxtkn run --verify "add a dark mode switch to the settings page"
Exit code
Your setup
Rules, skills, commands, tools and editors from other AI coding tools carry over, and each one stays under the same approval rules.
Reads AGENTS.md and .cursor rules as they are. A rule can always apply, match certain files, or apply only when you @name it.
SKILL.md folders made for other tools work without changes. Switch any skill off per project.
Type / to see every command. Write your own workflows, or import the ones from your .claude and .cursor folders.
Connect MCP servers over stdio or HTTP. Each tool asks for approval, and the server stops with the run.
Run your own programs before a command or after an edit, for example to block a commit until the tests pass.
Copy the Cursor rules, Claude Code skills and commands, MCP servers, hooks and Copilot instructions into Maxtkn. You see the exact text first, keys are never copied and your original files stay as they are.
Add agents, teams, rules, skills, commands, MCP servers and hooks in one step from a folder, a git address or a file. Anything that could run a program is off until you tick it, and uninstalling removes exactly what it added.
Ctrl+K finds actions, threads, files, commands and skills by a few letters. Type @ in the message box to point at a file, folder, thread, skill, agent or document.
Choose Light, Dark or match your device, turn on high contrast, or reduce motion. Colours are checked against the WCAG contrast rules, and a screen reader is told when the team needs you or finishes.
Presets for OpenRouter, Ollama, LM Studio, Gemini, Mistral, Groq and Azure, or your Claude Code or Codex sign-in. Choose which models show in the chat menu, see a usage dashboard, and set limits that warn at 80% and stop at 100%.
Open any file, or a change at its first changed line, in VS Code, Cursor, Zed, JetBrains and more.
A system notification when a run needs you, has a plan ready, finishes or fails. By default only when you are away.
Early access
Maxtkn is in early access and built on Windows today. Leave your email and choose a platform; we will write once, when a build is ready.
FAQ
No. Maxtkn is built for people who describe what they want and review results. The code is still one click away if you are curious, but you never have to read it.
After a run, Maxtkn runs your project checks and any acceptance checks you wrote in plain sentences, can open the app in a browser, and lists the evidence in the walkthrough. A check that could not run is reported as not run, never as a pass.
Anthropic Claude, any OpenAI-compatible API (OpenAI and similar), presets for OpenRouter, Gemini, Mistral, Groq and Azure OpenAI, and local servers such as Ollama or LM Studio. You add your own key in Settings, or use your Claude Code or Codex plan.
The app itself runs on your computer, but models need a connection unless you point Maxtkn at a local model server such as Ollama or LM Studio.
Yes, within limits. Goals, scheduled tasks and the command line run without you, with a time limit and a money limit for priced models. They never wait for an approval and never grant one, so anything risky is denied and listed in the report. Scheduled tasks run while Maxtkn is open.
Yes. Maxtkn reads AGENTS.md and .cursor rules, SKILL.md skill folders, slash commands from .claude and .cursor, and MCP servers. A folder that brings its own hooks or tools asks for your trust once, and lists exactly what would run.
On your computer. Projects are ordinary folders you choose, your keys are kept in the Windows Credential Manager, and your history and settings are stored in a local database. Model requests go directly from the app to the provider you configured.
Agents are limited to the project folder, paths that escape it are rejected, and commands are checked before they run. Access modes and permission levels decide what needs your approval, and risky commands always ask first. Every file change can be reverted, and you can rewind the project. By default this is a guard, not a virtual machine: you can switch on running commands in a container or WSL sandbox, and should keep Ask first for work you are unsure about.
The desktop app is built and tested on Windows today. macOS and Linux builds are not available yet. Join the early-access list and choose your platform to hear when they are.
Pricing is not announced yet. Join the early-access list and we will write to you when it is. Maxtkn does not resell model access. Model usage is billed by the provider whose key you use, and local models have no usage cost. Maxtkn shows an estimated cost for each run and lets you set limits.
Yes. The message box shows an estimate such as "About 40 cents, around 6 minutes" worked out from your own recent runs, or a labelled guess when there is no history yet. A finished run says what it took, and day and month budgets warn at 80% and stop at 100%.
Connect your own model and try the full workflow on a real project.