Aider is one of the best open-source AI coding tools available. It runs in your terminal, connects to models like GPT-4o, Claude, or Gemini, and pair-programs with you on real codebases. It reads your repo, proposes edits, applies them, and runs tests — all through a conversational CLI interface.
The problem: Aider sessions aren’t always quick. Ask it to refactor a module, implement a feature across multiple files, or debug a failing test suite, and you’re looking at 10-30+ minutes of back-and-forth. During that time, you’re tethered to your terminal. Step away, and you miss context, prompts, or error states that need your input.
MobileCLI fixes this by streaming any terminal session — including Aider — to your phone in real time.
What Aider Does (Quick Overview)
If you haven’t used Aider, here’s the short version. It’s a terminal-based AI pair programmer. You run it inside a git repo, it reads your codebase into context, and you have a conversation where you ask it to make changes. It edits files directly, creates commits, and can run shell commands.
Key characteristics relevant to mobile monitoring:
- Long-running sessions. Complex tasks involve multiple rounds of edits, test runs, and iterations. You start a task and wait.
- Interactive prompts. Aider asks for confirmation before applying certain changes or running commands. If you’re not there to respond, it blocks.
- Terminal-native. There’s no web UI or mobile app. It’s a CLI tool that expects you at a terminal.
- Rich output. Diffs, file listings, test output, model reasoning — there’s a lot to follow visually.
This interaction pattern — autonomous work punctuated by moments that need human attention — is exactly what MobileCLI was built for.
The Problem with Long-Running Aider Sessions
You ask Aider to implement a new API endpoint. It starts reading relevant files, proposes a plan, begins editing route handlers, models, and tests. Five minutes in, it hits an ambiguity — should it use middleware A or middleware B? It asks you. But you’re making coffee.
Ten minutes later you come back. Aider has been sitting idle, waiting. You answer, it resumes. Three minutes later, the tests fail. It proposes a fix and asks if you want to apply it. You’re on a call.
This pattern repeats. The actual compute time might be 15 minutes, but the wall-clock time is 45 because of human latency. Every context switch — walking back to your desk, reading the backlog, reorienting — costs time and focus.
The alternative is sitting there watching the terminal for the entire session. Which works, but it defeats the purpose of having an AI agent handle the work.
How MobileCLI Solves This
MobileCLI is a self-hosted Rust daemon that wraps any terminal process, captures its output, and streams it to the MobileCLI iOS app over WebSocket. You get:
- Live terminal output on your phone, rendered as a real terminal
- Push notifications when the process needs your attention
- File browsing to check what Aider has changed
- Input controls to respond to prompts without going back to your desk
It works with any terminal-based tool. No plugins, no API integrations, no configuration files. If it runs in a terminal, MobileCLI can stream it.
Setup: Aider + MobileCLI in 3 Minutes
1. Install MobileCLI
curl -fsSL https://mobilecli.app/install.sh | bashThis installs the mobilecli daemon on your workstation. It’s a single Rust binary — no Docker, no dependencies, no background services.
2. Install Aider (if you haven’t)
pip install aider-chatOr however you prefer to install it. MobileCLI doesn’t care about the installation method — it just wraps the process.
3. Run Aider Through MobileCLI
Instead of running aider directly, prefix it with mobilecli:
mobilecli aiderThat’s it. MobileCLI spawns Aider inside a PTY, starts capturing output, and displays a QR code in your terminal.
You can pass any Aider flags through:
mobilecli aider --model claude-3.5-sonnet --auto-commitsmobilecli aider --model gpt-4o --map-tokens 20484. Connect Your Phone
Open the MobileCLI iOS app and scan the QR code. Your phone connects directly to the daemon on your local network. No cloud relay, no third-party servers. The connection is peer-to-peer.
You’ll see Aider’s terminal output streaming live on your phone. From here, you can:
- Watch Aider work through your task in real time
- Respond to prompts and confirmation requests
- Browse files in your project directory to verify changes
- Receive push notifications when Aider is waiting for input
5. Walk Away
Start a task, put your phone in your pocket, and go about your day. When Aider needs you, your phone buzzes. Tap the notification, review the context, respond, and it continues.
What This Looks Like in Practice
Here’s a realistic workflow:
You (at desk): mobilecli aider
You (at desk): "Add rate limiting to the /api/users endpoint"
You (walking away): [puts phone in pocket]
Aider: [reads route files, middleware, tests...]
Aider: [proposes adding rate-limit middleware]
Phone: 🔔 "Aider is waiting for input"
You (on phone): [reviews proposed changes, types "yes"]
Aider: [applies changes, runs tests]
Aider: [tests pass, creates commit]
Phone: 🔔 "Aider session completed"The total active time you spent: maybe 30 seconds reviewing and approving. The rest was Aider working autonomously while you did something else.
Compared to Other Approaches
tmux + SSH from phone: Works, but mobile SSH clients are painful for reading dense terminal output. No push notifications. You have to manually check. The terminal rendering on a phone screen with a standard SSH client isn’t designed for readability.
Screen sharing / VNC: Heavy, slow, burns battery, requires a VPN or port forwarding. You’re streaming pixels instead of terminal state. Massive overkill for “did Aider ask me something?”
Just check periodically: The most common approach, and the worst. You either check too often (wasting time) or too rarely (Aider sits idle). Push notifications solve this completely.
Which Aider Workflows Benefit Most
Not every Aider session needs mobile monitoring. Quick one-off edits where you’re at your desk? Just use Aider directly.
Mobile monitoring shines for:
- Large refactors across multiple files where Aider iterates for 20+ minutes
- Test-driven workflows where Aider writes code, runs tests, fixes failures, and repeats
- Multi-model sessions where you’re running Aider with an architect model and want to review its high-level plan before it starts editing
- Overnight or background tasks where you kick off a big change and want to be notified when it’s done or stuck
- Shared workstation setups where you start Aider on a remote dev server and monitor from wherever you are
Technical Notes
MobileCLI captures raw PTY output, which means everything Aider renders — diffs, syntax-highlighted code blocks, progress indicators, token counts — shows up on your phone exactly as it would in your terminal.
The daemon is open source (MIT licensed) and available on GitHub. The iOS app is a Pro product — $20/year or $30 lifetime.
Since MobileCLI wraps the terminal process generically, it automatically works with any Aider version, any model backend, and any Aider configuration. There’s nothing Aider-specific to configure or maintain.
Getting Started
# Install MobileCLI
curl -fsSL https://mobilecli.app/install.sh | bash
# Run Aider through MobileCLI
mobilecli aider
# Scan the QR code with the iOS app
# DoneThree commands. No config files. Your Aider sessions are now on your phone.