New — multi-agent workflows

Describe it.Your agent ships it.

Metroit X turns plain-English into real code, commands, and workflows — run on your own machines. Connect every machine you own and drive them all from one screen — even from your phone.

Or for Linux macOS soon
Free — no card All your machines, one screen Even from your phone

Windows may show a "Windows protected your PC" notice on first install — that's normal for a new app. Click More infoRun anyway to continue.

14 tests passedin 2.3s
laptop · desktop · phone
Claude · Gemini · GPTrouted automatically
3 files editedauth/jwt_handler.py +2

Use the models you already know — pick per task, or let it route

GeminiClaudeGPTOpenRouter
Agentic coding, on your own machine

Not a chatbot. An agent that does the work.

Metroit X reads your codebase, plans, executes, and verifies — with proper guardrails. No more copy-pasting stack traces.

Plain English to real code

Describe what you want — the agent writes the files, runs the commands, and wires it together. Across your whole project, not one snippet at a time.

$ add JWT auth to my Flask app
→ reading project structure…
→ creating auth/jwt_handler.py
→ patching routes.py · adding deps
→ running pytest tests/auth/
✓ 14 passed in 2.3s
$

Every machine, one screen

Connect your laptop, desktop, and work PC at once. Run them in parallel or switch between them from a single dashboard — all your machines, controlled from one place.

Work from anywhere

A lightweight agent links each machine to the cloud. No laptop in your bag — fire off work and check the results right from your phone.

Self-healing workflows

When a command fails, the agent reads the error, fixes the cause, and retries — automatically.

You stay in control

Before any risky command or file write, the agent pauses and asks — approve once, approve for the session, or deny.

Built for every kind of dev

However you ship, Metroit X fits in.

Ship side projects on weekends, not weeknights.

When it's just you, every hour spent on boilerplate is an hour stolen from the actual idea. Metroit X handles the plumbing so you can stay in the fun part.

  • Free tier covers most side projects — connect 1 machine, weekly usage, all core features.
  • No API keys to manage — Claude, GPT, DeepSeek, and more — included in every plan, ready to use.
  • One-shot scaffolds — “set up a Next.js project with auth + Stripe” → ready in minutes.
weekend-ai-app
3 files · Next.js · last touched 2h ago
solo
side-cli-tool
12 files · Python · yesterday
solo
chrome-extension
5 files · TypeScript · 3 days ago
solo
Your ~/projects — mirrored to Metroit X cloud, on-demand only.

From idea to shipped in three steps

No new IDE to learn. Connect, ask, done.

01

Connect your machine

Install the agent and pair it in one click. Your workspace shows up in the dashboard instantly.

02

Ask in plain language

“Add JWT auth to my Flask app”, “fix this failing test”, “deploy to staging.” No special syntax.

03

Agent builds & verifies

It reads your code, makes changes, runs commands, checks the result — and asks before anything destructive.

Try it yourself

Type a prompt. See what happens.

Live preview of what the agent does — pick a preset or write your own.

metroit-x ~ live preview
Hit Run or pick a preset to see the agent work…
tool_call → ok
writing file…
Real-time streaming

Watch every step as it happens.

Tool calls, file edits, command output — streamed live over WebSocket. No black box, no "trust me bro."

  • Multi-agent parallelismSpin up sub-agents for parallel work — one fixes tests, another writes docs, a third deploys.
  • Route across modelsCheap models for routing & classification, powerful ones for actual building. Spend smart.
  • MCP-nativePlug in any Model Context Protocol server — databases, browsers, custom tools. The agent uses them like a senior dev would.
Fits your workflow

Works with the stack you already use.

Git, package managers, containers, your terminal — and any MCP server you plug in. Metroit X runs the tools your project already uses.

GitHub
Git
npm
Docker
Terminal
MCP servers
Simple, honest pricing

Start free. Upgrade when you ship more.

No per-seat traps, no hidden fees. Cancel anytime.

Free

For trying it out and small side projects.

Free
Start free
  • Connect 1 machine
  • All core agent features
  • Multi-model (Gemini, Claude, GPT, Kimi)
  • Weekly usage for a few projects
  • Community support

Max

For power users who code with agents all day.

₹4,999/month
Upgrade to Max
  • Connect up to 10 machines
  • Highest usage limits
  • Priority routing to fastest models
  • All Pro features included
  • Higher daily safety cap
  • Priority support
Security & trust

Your code, your secrets, your rules.

Your code stays on your machine. Permission prompts before risky actions, full audit logging, and secrets scrubbed from logs — security built in, not bolted on.

Runs on your machine

The agent lives on your machine. It only reads files it needs, only when you ask, and never sends your project tree wholesale.

Permission gates

Every risky action — file delete, package install, git push — pauses for explicit approval. Auto-approvals are session-scoped.

Audit log

Every prompt, tool call, file edit, and command is logged with timestamps, so you can always see exactly what the agent did.

Encrypted in transit

All traffic runs over HTTPS. Auth is JWT with revocation and bcrypt-hashed passwords, and secret-shaped values are scrubbed from logs — so keys never leak.

Runs on your machinePermission promptsAudit loggingJWT auth + revocationbcrypt password hashingSecrets scrubbed from logs

Common questions.

Does my code leave my machine?
Only the snippets the agent needs to reason about — and only when you've approved a request. The agent runs locally and you can scope what's shared per-project.
Which models can I use?
Gemini, Claude, GPT, plus anything available through OpenRouter. Pick per task or let the router decide based on cost and complexity.
Can it run real terminal commands?
Yes — but always with permission. The first time a session tries something risky (file delete, git push, package install), it pauses and asks. You can allow once, allow for the session, or deny.
What happens when the agent gets stuck?
It reads the error, attempts a fix, and retries — up to a configurable limit. If it can't recover, it stops and shows you exactly what it tried, so you can take over without losing context.
Is there a free tier?
Yes — Free includes 1 machine, weekly usage for small projects, and every core agent feature. No card required.

It started on a 3–4 hour drive out of town. I wanted to keep working, but my laptop wasn't with me — just my phone, and hours about to go to waste. So I built Metroit X: connect your machine once, then run real work from anywhere. Your agent stays linked to your projects — so you can ship code, run commands, and check the results right from your phone. No laptop in your bag, no time wasted. That car ride is exactly who Metroit X is for.

AT
Aniket T.Aniket ThakurFounder, Metroit X

Ship your next feature by describing it.

Connect your machine, start free, and let Metroit X handle the typing. You stay in control of every step.