Our Story — We Built the Terminal We Couldn't Find Anywhere Else

We're a small team that ships fast. In six months, we built 12 projects — all with AI coding assistants like Claude and GPT. We loved the speed. But one thing kept slowing us down: we couldn't leave our desks. Every time an AI agent asked a question, needed a permission, or hit an error — we had to be right there, staring at the terminal.

So we built Remocode. Here's exactly how we use it, every single day.

Step 1: Open Remocode — Four Terminals, Ready to Go

You download, open, and you're greeted with four terminal panes. cd into each project folder — that's it. Remocode auto-saves every path. Next time you launch, each pane opens exactly where you left it.

‹›

1 / 2— click arrows

Step 2: Always Know Where You Are

When you're running 4 Claude agents at once, every pane looks the same — a wall of code and AI output. You can't just run ls mid-session without interrupting the agent. Remocode shows the working directory for each pane at all times.

Step 3: Add Your AI Key — Get a Second Opinion

Open settings (Cmd+Shift+A → gear icon), drop in your API key — Claude, GPT, Gemini, or local Ollama. Now you can chat with AI about your terminal output. Ask it to audit code, review what your agents are doing, or get an expert second opinion.

‹›

1 / 2— click arrows

Step 4: Set Up Telegram — Two Minutes, Tops

Open @BotFather on Telegram, create a bot, paste the token into Remocode. Then message @userinfobot to get your Chat ID and paste that too. That's all — you're now connected to your terminals from anywhere in the world.

‹›

1 / 4— click arrows

Three steps in the settings:

• ●Create bot via @BotFather → copy token
• ●Message @userinfobot → copy Chat ID
• ●Paste both in Remocode Settings → Provider tab

Step 5: Type "status" — See Everything at Once

Your first Telegram message is always status. It tests the connection and immediately gives you a summary of all four projects — what each agent is doing, where they're stuck, what's passing. One message, full picture.

you > status

● Pane 1 (api-server): Auth module done, 12 tests passing
● Pane 2 (mobile-app): Stuck — asking about navigation lib
● Pane 3 (landing-page): Building hero section, no issues
● Pane 4 (chrome-ext): Idle since 14:22

Step 6: Focus with "via" — The Killer Feature

You see Pane 2 is stuck. Type via mobile-app to connect directly to that terminal. Now it's bidirectional — terminal output streams to Telegram, your messages go straight to the terminal.

First thing you do? reply — it fetches the latest AI output so you can see what's happening. Read it, type your answer, hit submit. The agent continues. All from your phone.

‹›

1 / 2— click arrows

you > via mobile-app
Connected to Pane 2

you > reply
ai  > Claude is asking: "Which navigation library?
      1. React Navigation
      2. Expo Router
      3. React Native Nav"

you > 2
you > submit
✓ Sent to terminal

If something seems off — maybe your command didn't go through — use peek to check if there's leftover unsubmitted input sitting in the terminal.

Step 7: The Loop — Reply, Answer, Reply

This is where Remocode really clicks. You're at a coffee shop, on a walk, wherever. The loop is simple: reply to check what's happening, answer if the AI is stuck, reply again to watch it work. Auto-notifications ping you when a task finishes. In between — reply is how you stay in the flow.

you > reply
ai  > Building Expo Router layout... 3 files created

     15 min later...

you > reply
ai  > Navigation complete. 8/8 tests passed.
      Should I add deep linking? (y/n)

you > y
you > submit

It becomes muscle memory. Reply. Answer. Reply. That's the rhythm.

Step 8: Auto-Scroll — Stop Fighting the Terminal

Here's a pain every AI coder knows: Claude writes a long response, the terminal redraws line by line, and the scroll jumps to the top. You're yanked away from the output you were reading. Over and over.

The ⇣ button in each pane's title bar fixes this. Click it once — it locks your view to the bottom. Every new line of output scrolls you down automatically. Click again to unlock, so you can scroll up and read history without being yanked back.

Step 9: Auto-Yes — Stop Clicking "Proceed?" Forever

Claude Code asks "proceed?" about 50 times per session. Every prompt breaks your flow. The Y button in the title bar fixes this.

Click Y once — it turns green. Remocode now scans every 2 seconds for Yes/No prompts and auto-selects Yes. Zero AI cost. Pure pattern matching. Your agent keeps moving without waiting for you.

Right-click the Y button to open the Auto-Yes Log — a full list of everything it auto-approved, with timestamps and the original question.

Step 10: AI Supervisor — Let AI Manage Your AI

Auto-Yes is great for simple approvals. But what about smarter decisions — approve safe file edits, reject rm -rf, escalate new dependencies?

That's the 🧠 brain icon. Click it for the first time and it opens the Brief Editor — write your project rules in plain English:

## Project: E-commerce API
- Approve: file reads, test runs, npm install
- Reject: database migrations, rm commands, anything /prod
- Escalate: new dependencies, API changes

Save the brief, and the supervisor starts scanning every 2 seconds. It reads your terminal, decides what to do, and acts — approve, reject, respond, or escalate to you.

Right-click the 🧠 button to open the Decision Log — every action the supervisor took, its reasoning, risk level, and cost. Full transparency.

‹›

1 / 2— click arrows

Step 11: Test Write — Try Before You Trust

Not sure if the supervisor will make the right call? The ✍ button lets you test it.

Click ✍ once — it reads the last 50 lines of your terminal, sends them to AI, and shows you the decision. The button changes to ⏳ while thinking, then flashes green (approved) or red (escalated) for 4 seconds.

It's a one-shot test. No auto-scanning, no loop — just "what would the supervisor decide right now?" Use it to tune your brief before enabling the full supervisor.

Step 12: Pick Your Autonomy Level

Y (Auto-Yes) and 🧠 (Supervisor) can't run at the same time on the same pane — enabling one disables the other. This is by design.

• ●Use Y when you fully trust the agent — "just keep going, approve everything"
• ●Use 🧠 when you want smart decisions — "approve safe stuff, reject dangerous stuff, escalate what you're unsure about"
• ●Use ✍ anytime to spot-check what the supervisor would decide

Mix and match across panes: Y on your trusted pane, 🧠 on the risky one. Each pane is independent.

Helpful Commands — Your Cheat Sheet

Forgot a command? Just type help in Telegram — you'll get the full list, right there in the chat. Don't remember your terminal names? Type via anything — Remocode will show you every available terminal. No manual needed.

Terminal Control:

• ●via <name> — jump into a bidirectional terminal session
• ●enter / submit — send the Enter key
• ●peek — check for leftover unsubmitted input
• ●reply — fetch the latest AI output

AI Commands:

• ●status — get a progress report across all terminals
• ●status <name> — report on one specific terminal
• ●audit <name> — run a security audit

Pro tip: Type via with any wrong name to instantly see all your terminal names. It's the fastest way to orient yourself.

Built by Developers, for Developers

We built Remocode because we needed it — and we're shipping new features every week based on real usage. If something's broken, missing, or could be better, we want to hear it. No support tickets. No chatbots. Just developers talking to developers.

Drop us a line at developer@trini-ai.com — whether it's a bug report, a feature request, or just to say what's working. Every email goes straight to the team that builds this thing.

Start with a 7-day Pro trial — no credit card required. Download Remocode, set up Telegram in 2 minutes, and never be chained to your desk again.